Features:
* Block incremental backup. (Reviewed by John Morris, Stephen Frost, Stefan Fercot.)
* SFTP support for repository storage. (Contributed by Reid Thompson. Reviewed by Stephen Frost, David Steele.)
* PostgreSQL 16 support. (Reviewed by Stefan Fercot.)
Improvements:
* Allow page header checks to be skipped. (Reviewed by David Christensen. Suggested by David Christensen.)
* Avoid chown() on recovery files during restore. (Reviewed by Stefan Fercot, Marcelo Henrique Neppel. Suggested by Marcelo Henrique Neppel.)
* Add error retry detail for HTTP retries.
Documentation Improvements:
* Add warning about using recovery type=none. (Reviewed by Stefan Fercot.)
* Add note about running stanza-create on already-created repositories.
The prior timeouts were a bit aggressive and were causing timeouts in the Azure tests. There have also been occasional timeouts in other storage drivers.
The performance of CI environments is pretty variable so increased timeouts should make the tests more stable.
Double spaces have fallen out of favor in recent years because they no longer contribute to readability.
We have been using single spaces and editing related paragraphs for some time, but now it seems best to update the remaining instances to avoid churn in unrelated commits and to make it clearer what spacing contributors should use.
These were intended to allow the block list to be scanned without reading the map but were never utilized. They were left in "just in case" and because they did not seem to be doing any harm.
In fact, it is better not to have the block numbers because this allows us set the block size at a future time as long as it is a factor of the super block size. One way this could be useful is to store older files without super blocks or a map in the full backup and then build a map for them if the file gets modified in a diff/incr backup. This would require reading the file from the full backup to build the map but it would be more space efficient and we could make more intelligent decisions about block size. It would also be possible to change the block size even if one had already been selected in a prior backup.
Omitting the block numbers makes the chunking unnecessary since there is now no way to make sense of the block list without the map. Also, we might want to build maps for unchunked block lists, i.e. files that were copied normally.
The chown() was already skipped on the files restored from the repository but the same logic was not applied to the generated recovery files, probably because chown'ing a few recovery files does not have performance implications. Use the same logic for recovery files to determined if they need to be chown'd.
Ultimately this behavior is pretty hard to test, so add a fail safe into the Posix driver that will skip chown if the permissions are already as required.
These checks cause false negatives for page checksum verification when the page is encrypted because pd_upper might end up as 0 in the encrypted data. This issue is rare but reproducible given a large enough cluster.
Make these checks optional, but leave them enabled by default.
Centralize the code to allow it to be used in more places and update the protocol/server module to use the new code.
Since the time measurements make testing difficult, also add time and errorRetry harnesses to allow specific data to be used for testing. In the case of errorRetry, the production behavior is turned off by default during testing and only enabled for the errorRetry test module.
Ubuntu 18.04 will be EOL before the next release, so update to the oldest available Debian version.
Also fix one incorrect return value type, a test cast, and adjust some test timeouts.
A lot of these are left over from when object interfaces required allocations (changed in f6e30736 and 9ca9c8e4). Others are likely copy/paste errors.
This saves some space in the mem context and makes it clear that no allocations will be made.
The result is not intended to be freed directly so this makes memory tracking more accurate. Fix a few places where memory was leaking after a call to zNewFmt().
Also update an assert to make it clearer.
Eliminate the boilerplate of declaring this and assigning memory to it, which is the same for the vast majority of object creations.
Keep the old version of the macro as OBJ_NEW_BASE_BEGIN() for a few exceptions in the core code and (mostly) in the tests.
As in f6e30736, make the interface object the parent of the driver object rather than the interface being allocated directly in the driver object. Allow exceptions to this general rule for objects that need to retain ownership of their interfaces.
Each call to lockAcquireP() passed enough information to initialize the lock system. This was somewhat inefficient and as locks become more complicated it will lead to more code duplication. Since a process can only take one type of lock it makes sense to do most of the initialization up front.
Also reduce the log level of lockRelease() since it is only called at exit and the lock will be released in any case.
This minor optimization automatically clears the limit flag when limit is set to allocated size. This has no impact on the current code but will simplify a future commit where a conditional bufLimitClear() is being used.
Bug Fixes:
* Skip writing recovery.signal by default for restores of offline backups. (Reviewed by Stefan Fercot. Reported by Marcel Borger.)
Features:
* Block incremental backup (BETA). (Reviewed by John Morris, Stephen Frost, Stefan Fercot.)
Improvements:
* Keep only one all-default group index. (Reviewed by Stefan Fercot.)
Documentation Improvements:
* Add explicit instructions for upgrading between 2.x versions. (Contributed by Christophe Courtois. Reviewed by David Steele.)
* Remove references to SSH made obsolete when TLS was introduced.
Output a manifest in text or JSON format. Neither format is complete but they cover the basics.
In particular the manifest command outputs the complete block restore when the filter option is specified and the block delta when the pg option is also specified. This is non-destructive so it is safe to execute against a running cluster.
Bug Fixes:
* Remove the distinction between maps where super block size is equal to block size and maps where they are not. In practice, maps with equal blocks are now rare and most of the optimizations can be applied directly to super blocks where the blocks are equal. This fixes a bug where a map that was created with equal size blocks and then converted to differing block sizes would generate an invalid map.
* Free reads during restore to avoid running out of file handles.
Improvements:
* Store super block sizes in the block map. This allows the final block size to be removed from the block list and provides a more optimal restore and better potential for analysis.
* Always round the super block size up to the next block size. This makes the number of blocks per super block more predictable.
* Allow super block sizes to be changed at will in the map. The first case for this is to store the reduced super block size required when the last super block is short but it could be used to dynamically change the super block size to optimize compression.
* Store a block count rather than a list of blocks in a super block. Blocks must always be sequential, though there may be an offset to the first block in a super block. This saves 11-14% on space for checksum sizes 6-7.
* In the case that all the blocks for a super block are present, and there is no offset, the block size is omitted.
This allows options to be marked as beta, which will require that the --beta option be supplied to prevent accidental usage of a beta feature.
The online and command-line documentation also show warnings when options are beta.
Block sizes are incremented when the size of the map becomes as large as a single block. This is arbitrary but it appears to give a good balance of block size vs map size.
The full backup super block size is set to minimize loss of compression efficiency since most blocks in the database will likely never be modified. For diff/incr backup super blocks, a smaller size is allowable since only modified blocks are stored. The overall savings of not storing unmodified blocks offsets the small loss in compression efficiency due to the smaller super block and allows more granular fetches during restore.
As calculated this size is not correct since it does not include the parts of prior block incrementals that are required to make the current block incremental valid. At best this could be approximated and the resulting values might be very confusing.
For now, at least, exclude this metric for block incremental backups.
xxHash is significantly faster than SHA-1 so this helps reduce the overhead of the feature.
A variable number of bytes are used from the xxHash depending on the block size with a minimum of six bytes for the smallest block size. This keeps the maps smaller while still providing enough bits to detect block changes.
Small blocks sizes can lead to reduced compression efficiency, so allow multiple blocks to be compressed together in a super block. The disadvantage is that the super block must be read sequentially to retrieve blocks. However, different super block sizes can be used for different backup types, so the full backup super block sizes are large for compression efficiency and diff/incr are smaller for retrieval efficiency.
Forks may update pg_control version or WAL magic without affecting the structures that pgBackRest depends on.
This option forces pgBackRest to treat a cluster as the specified version when it cannot be automatically identified.
When restoring an offline backup on PostgreSQL >= 12, skip writing recovery.signal by default since this will error if the backup was made with wal_level=minimal. If the user explicitly sets the type option to something other than none, then write recovery.signal as usual since it is possible to do Point-In-Time-Recovery from an offline backup as long as wal_level was not minimal.
Raw encryption was already being used for block incremental. This commit adds raw compression to block incremental where possible (see da918587).
Raw compression/encryption is also added to bundling for a backup set when block incremental is enabled on the full backup. This prevents a break in backward compatibility since block incremental is not backward compatible.
Raw format saves 12 bytes of header for gzip and 4 bytes of checksum for lz4 (plus CPU overhead). This may not seem like much, but over millions of small files or incremental blocks can really add up. Even though it may be a relatively small percentage of the overall backup size it is still objectively a large amount of data.
Use raw format for protocol compression to exercise the feature.
Raw compression format will be added to bundling and block incremental in a followup commit.
Make the interface object the parent of the driver object rather than the interface being allocated directly in the driver object.
The prior method was more efficient when mem contexts had a much higher cost. Now mem contexts are cheap so it makes more sense to structure the objects in a way that works better with mem context auditing. This also means the mem context does not need to be stored separately since it can be extracted directly from the interface object.
There are other areas that need to get the same improvement before the specialized objMoveContext() and objFreeContext() functions can be removed.
This flag does not currently affect restore behavior but it will in an upcoming commit. Set the flag here to simplify the test diff in the upcoming commit.
It is possible for a group index to be created for an option that is later found to not meet dependencies. In this case all values would be default leading to a phantom group, which can be quite confusing.
Remove group indexes that are all default (except the final one) and make sure the key for the final all default group index is 1.
The code is not completely reflowed yet so there are some cases that uncrustify will not catch. The formatting will be improved over time.
Some block of code require special formatting so have been surrounded with the {uncrustify-off}/{uncrustify-on} markers. These exceptions should be kept to a minimum.
Add --code-format (to reformat code) and --code-format-check (to check formatting) to test.pl.
Add a CI test that will check code formatting. Code must be correctly formatted before it can be merge to integration.
Add documentation to the coding standards for code formatting.
uncrustify has been configured to be as close to the current format as possible but the following changes were required:
* Break long struct initializiers out of function calls.
* Bit fields get extra spacing.
* Strings that continue from the previous line no longer indented.
* Ternary operators that do not fit on a single line moved to the next line first.
* Align under parens for multi-line if statements.
* Macros in header #if blocks are no longer indented.
* Purposeful lack of function indentation in tests has been removed.
Currently uncrustify does not completely reflow the code so there are some edge cases that might not be caught. However, this still represents a huge improvement and the formatting can be refined going forward.
Support code for uncrustify will be in a followup commit.
Bring the format(printf) attribute in line with the FN_NO_RETURN and FN_INLINE_ALWAYS macros.
This is simpler to read and can be customized for different compilers.
stackTraceToZ() was split this way in c8264291 to allow complete coverage. 0becb6da added a shim to improve coveage but missed simplifying the function.
Improvements:
* Remove support for PostgreSQL 9.0/9.1/9.2. (Reviewed by Stefan Fercot.)
* Restore errors when no backup matches the current version of PostgreSQL. (Contributed by Stefan Fercot. Reviewed by David Steele. Suggested by Soulou.)
* Add compress-level range checking for each compress-type. (Reviewed by Stefan Fercot. Suggested by gkleen, ViperRu.)
Documentation Improvements:
* Add warning about enabling "hierarchical namespace" on Azure storage. (Reviewed by Stefan Fercot. Suggested by Vojtech Galda, Pluggi, asjonos.)
* Add replacement for linefeeds in monitoring example. (Reviewed by Stefan Fercot. Suggested by rudonx, gmustdie, Ivan Shelestov.)
* Clarify target-action behavior on various PostgreSQL versions. (Contributed by Chris Bandy. Reviewed by David Steele, Anton Kurochkin, Stefan Fercot. Suggested by Anton Kurochkin, Chris Bandy.)
* Updates and clarifications to index page. (Reviewed by Stefan Fercot.)
* Add dark mode to the website. (Suggested by Stephen Frost.)
Running valgrind and backtrace together has been causing tests to timeout in CI, mostly likely due to limited resources. This has not been a problem in normal development environments.
Since it is still important to run backtraces for debugging, split the u22 test that was doing all this work to run coverage and backtrace together and valgrind-only as a separate test. As a bonus these tests run faster separately and since they run in parallel the total execution time is faster.
The primary goal of the block incremental backup is to save space in the repository by only storing changed parts of a file rather than the entire file. This implementation is focused on restore performance more than saving space in the repository, though there may be substantial savings depending on the workload.
The repo-block option enables the feature (when repo-bundle is already enabled). The block size is determined based on the file size and age. Very old or very small files will not use block incremental.
The callbacks in iniLoad() made the downstream code more complicated than it needed to be so use an iterator model instead.
Combine the two functions that were used to load the ini data to remove code duplication. In theory it would be nice to use iniValueNext() in the config/parse module rather than loading a KeyValue store but this would mean a big change to the parser, which does not seem worthwhile at this time.
It is possible for functions to accidentally leak child contexts into the calling context, which may use a lot of memory depending on the use case and where it happens.
Use the function return type to determine what should be returned and error when something else is returned. Add FUNCTION_AUDIT_*() macros to handle exceptions.
This checking is only performed during unit tests on the code being covered by the specific unit test.
Note that this does not work yet for memory allocations, i.e. memNew(). These are pretty rare so are not as much of an issue and they can be added in the future.
Allocating memory made these functions simpler but it meant that memory was leaking into the calling context when logging was enabled. It is not clear that this was an issue but it seems that trace level logging could result it a lot of memory usage depending on the use case.
This also makes it possible to audit allocations returned to the calling context, which will be done in a followup commit.
Also rename objToLog() to objNameToLog() since it seemed logical to name the new function objToLog().
This fills in backtrace info at the bottom of the call stack when the stack trace is incomplete due to testing. This does not affect release builds, which is why it did not make the first cut, but it turns out to be useful for testing and barely changes the release code (when we do release this).
The recursion test in common/error was simplified because it would now return a very large trace.
The error detail should be output when the error is an assert (this part was working) or the log level is at least debug. In cases where log-level-console was at least debug but log-level-stderr was not the detail was lost.
Improve the range checking to output error detail to stderr when log-level-console is at least debug.
The libbacktrace feature has not been working since the move to meson because libbacktrace detection was not added to the meson build. Add libbacktrace to meson and improve the feature so that it can be compiled into release builds.
The prior implementation fetched line numbers with each stack trace push. Not only was this slow but it missed any functions that were not being tracked on our stack.
Instead just examine the backtrace when an error happens and merge it with the info we have on our stack. If the backtrace is not available then the output remains as before.
Also remove --backtrace from test.pl since the library is now auto-detected.
Leave this library out of the production build for now to give it a little time to shake out in testing.
When this code was migrated to C the unit tests were not included because there were more important priorities at the time.
This also requires some adjustments to coverage because of the new code location.
BUFFER_EXTERN() provides a clean way to create buffer constants.
Convert HASH_TYPE_SHA256_ZERO_STR to HASH_TYPE_SHA256_ZERO_BUF to be consistent with HASH_TYPE_SHA1_ZERO_BUF.
This should make it a little clearer what the variable (VR) macros are doing since the declaration/definition cannot both be set to extern (but functions can).
Splitting the variable macros out also allows them to be changed in the future with little churn, while changing the function macro creates a large amount of churn.
This is immediately useful because it will detect any extern'd functions or variables that are not being used. It also detects functions or variables that are declared but not defined.
If a FV/VR_EXTERN macro is missing it will be detected either because of a mismatch in the declaration/definition or because a new defined symbol will appear in the nm test.
Eventually the unity build will be used to create a more optimized pgbackrest binary but that will need to wait.
Similar to b9be4fa5, these functions are not used by the core code so move them to the build module. The new implementation is a little less efficient but that is much less of a worry in the build/test code.
Also remove regExpMatchSize() since it was not longer needed.
Neither of these functions were used by the core code. strReplace() is only used in the tests but it doesn't hurt to put it in build since the build code is not distributed.
This was done by checking the extension but it is possible to include a module that does not have a vendor or auto extension. Instead make it explicit that the module is included in another module.
Also change the variable from "include" to "included" to make it clearer what it indicates.
It is probably not a good idea to restore the latest backup when it was not made from the current PostgreSQL version. If there is no backup after a stanza-upgrade then replicas might be built with a prior version leading to failures.
Add an error in this case if the latest backup would be used, i.e. --set or --type=time/lsn is not specified.
The prior range checking was done based on the valid values for gz. While this worked it was a subset of what is available for lz4 and zst.
Allow the range to be specified for each compress-type. Adding this functionality to the parse module would be a better solution but that is a bigger project than this fix deserves, at least for now.
Calculate a checksum of the data stored in the repository when a file is transformed (e.g. compressed). This allows resume and verify to operate without needing to decompress/decrypt the data.
This can also be used to verify more complex formats such as block incremental and allow backups from the repository without needing to decompress the data to verify the checksum.
Add some basic encrypted tests to maintain coverage. These will be expanded in a future commit.
Manifest checksums were stored as hex-encoded strings due to legacy compatibility with Perl. Storing the checksums as binary in memory uses half the space and avoids many conversions.
There is no change to the on-disk manifest format which stores the checksum as a hex-encoded string.
Our new policy is to support ten versions of PostgreSQL, the five supported releases and the last five EOL releases. As of PostgreSQL 15, that means 9.0/9.1/9.2 are no longer supported by pgBackRest.
Remove all logic associated with 9.0/9.1/9.2 and update the tests.
Document the new support policy.
Update InfoPg to read/write control versions for the history in backup.info, since we can no longer rely on the mappings being available. In theory this could have been an issue after removing 8.3/8.4 if anybody was using a version that old.
This avoids constructs such as decodeToBin(encodeBase64, ...) which are confusing since decode and encode are used in the same function call. decodeToBin(encodingBase64, ...) makes it clearer what is happening.
The storageNewItrP() permissions test was running twice with the errorOnMissing flag set to false. Fix by setting to true for one test.
Also update the comments to be clearer about what the tests are doing and fix minor formatting.
This allows test backups to be run in other test modules.
It is likely that more logic will be moved here but for now this suffices to get test backups working in the restore module.
When loading prior manifests without the new reference list, the code failed to add the current backup to the reference list. Since the current backup is never explicitly referenced, building references from the file list was not sufficient to generate a complete list.
The main problem here was a bad test, fixed in 28f6604. This masked the issue and prevented it from being found. Now it is clear in the test that the current label is missing from the reference list.
Fix by adding the current label to the reference list if a reference list is not stored in the manifest.
Changing the label of a manifest that already had a label was not a good test and it ended up masking a bug where the current backup label was not being added to the reference list on manifest load, since manifestBackupLabelSet() added the label to the reference list. In fact, manifestBackupLabelSet() should never be called after a manifest load or even after the label has been set.
Add an assertion to prevent manifestBackupLabelSet() being called when the label is already set.
The bug exposed here will be fixed in a subsequent commit.
Hopefully this will make it a little clearer to the user what is wrong when they specify an indexed option without an index.
Also fix an ambiguous use of cfgParseOptionP(). The prior code worked in that it set prefixMatch = true but it was not very readable.
Bug Fixes:
* Fix memory leak in file bundle backup/restore. (Reviewed by John Morris, Oscar. Reported by Oscar.)
* Fix protocol error on short read of remote file. (Reviewed by Stephen Frost.)
Improvements:
* Do not store references for zero-length files when bundling. (Reviewed by Stefan Fercot.)
* Use more generic descriptions for pg_start_backup()/pg_stop_backup(). (Reviewed by Greg Sabino Mullane, David Christensen. Suggested by Greg Sabino Mullane.)
Test Suite Improvements:
* Update test.pl --psql-bin option to match command-line help. (Contributed by Koshi Shibagaki. Reviewed by David Steele.)
The option to specify the path to psql was shown in the command-line help as --psql-bin but the option was actually named --pgsql-bin.
Rename to match the help so they are consistent.
The magic in the header is only required so that command-line openssl will recognize the file as being encrypted. In cases where the encrypted data cannot be read with the command-line tool it makes sense to omit the header magic to save some space.
Unfortunately this cannot be enabled for file bundling because it would break backward compatibility. However, it should be possible to enable it for the combination of bundling and block incremental.
The prior code required coverage in the storage/remote module for all filters that could be used remotely.
Now the filter handlers are set at runtime so any filter list can be used with a remote. This is more flexible and makes coverage testing easier. It also resolves a test dependency.
Move the command/remote unit test near the end so it will have access to all filters without using depends.
This flag skips truncation when opening a file for write on drivers that support it, currently Posix and CIFS. This is convenient for cases where the file needs to be manipulated directly using the file descriptor. Using the file descriptor is not ideal and additional functionality should be added to the storage interface, but for now at least this avoids code duplication, especially on close which updates owners, the timestamp, syncs, etc.
The remote driver forbids no truncate because a file descriptor is never available for a remote storage write object.
Update two instances in the current code which benefit from this new functionality, but the primary reason for the change is to support more complex restore deltas in the upcoming block incremental feature.
If a remote file read was stopped before the read was complete or if an error occurred in the middle of the read then the protocol would end up in a bad state and produce this error:
ProtocolError: client state is 'data-get' but expected 'idle'
Prevent this by reading the rest of the file on close() or free() to leave the protocol in an idle state for the next command.
This was a possible issue for bundling because the amount to read is known in advance and therefore eof may not be reached. However, I was only able to reproduce this issue with unreleased code.
On error this issue would cause the original error to be lost. The process may still fail with this fix (if the error comes from another source) but hopefully we'll get better information about the original error.
The names were changed in PostgreSQL 15, so update the code and docs to make the naming more generic where needed to avoid using a version-specific name in the logs and documentation.
The prior result was hex-encoded, which is not optimal. This was legacy from the interface with Perl and then the JSON protocol. The new binary protocol natively supports binary so it makes sense to use it and convert to hex where needed.
A number of these hex conversions can now be removed but that will need to be handled in another commit.
This makes it more efficient to read/write (especially read) varint-128 to/from IO.
Update the Pack type to take advantage of the more efficient read and remove some duplicate code.
The reference list was previously built at load time from whichever references existed in the file list. This was sufficient since the list was for informational purposes only.
The block incremental feature will require a reference list that contains all prior backups, even those that are not explicitly referenced from the manifest. Therefore it makes sense to build and persist a manifest list rather than building it at load time.
This list can still be used for informational purposes, though it needs to be sorted since the list it sill built for older manifest versions and may not be in sorted order.
Add strLstFindIdx() to find references in the list.
The prior method was to check a combination of fields to determine if a file needed to be copied, delta'd, or resumed. This was complicated and ultimately imposed a limitation on the number of operations that could be performed.
Introduce copy, delta, and resume flags in the manifest to make it clearer which operations need to be performed and to reduce complex and duplicated logic.
This also allows zero-length bundled files to be completed during manifest build rather than later on during backup processing.
The prior manifestFileUpdate() function was pretty difficult to use since all the parameters had to specified. Instead, pass a ManifestFile struct that has all members set as needed.
When new struct members are added the manifestFileUpdate() call sites will still need to be reviewed, but this should make the process of adding members a bit simpler.
This appears to have been an oversight in 34d6495. Storing the reference is not really correct since the file is not stored in a prior backup. It also uses more space.
There is no real harm in storing the reference, since it is always ignored on restore, but the code is simpler if the zero-length files can be dealt with during the manifest and don't need additional handling later on. This is also an important part of some upcoming optimizations.
It is possible to log the bundle info correctly but the information is useless with the backup reference, which does not appear until later. For now just omit the bundle info so we are not logging something incorrect.
This test exposes a small logging issue. The bundle information for the matched delta on PG_VERSION is not correct. This issue will be fixed in the next commit.
The information stored in the manifest *is* correct so this bug is essentially cosmetic.
The current call site, manifestFileUnpack(), does not know the total buffer size but the buffer has always been maintained in memory so there should be no corruption. However, there are upcoming use cases where the buffer will be read from IO, the buffer size will be known, and additional sanity checking on buffer overruns will be valuable.
Also rename params to align better with cvtUInt64ToVarInt128().
Direct link creation via Posix functions has been moved to the Posix driver.
This change allows adding SFTP softlink creation in the SFTP driver using the standard interface.
Ninja produces quite a bit of output so error messages are often truncated by the default error/log buffers. Use large buffers in the test harness to capture the error even when there is a lot of output.
Ninja has introduced a --quiet option, but it is currently too new to be in any of our test distributions.
Bug Fixes:
* Fix incorrect time expiration being used for non-default repositories. (Reviewed by Stefan Fercot. Reported by Adam Brusselback.)
* Fix issue when listing directories recursively with a filter. (Reviewed by Stephen Frost. Reported by Efremov Egor.)
Features:
* Backup key/value annotations. (Contributed by Stefan Fercot. Reviewed by David Steele. Suggested by Adam Berlin.)
Improvements:
* Support --set in JSON output for info command. (Contributed by Stefan Fercot. Reviewed by David Steele. Suggested by Anton Kurochkin.)
* Update archive.info timestamps after a successful backup. (Reviewed by Stefan Fercot. Suggested by Alex Richman.)
* Move standby timeline check after checkpoint. (Reviewed by Stefan Fercot, Keith Fiske. Suggested by Keith Fiske.)
* Improve warning message on backup resume. (Suggested by Cynthia Shang.)
Documentation Improvements:
* Add absolute path for kill in pgbackrest.service. (Suggested by Don Seiler.)
While recursing and filtering, if the last entry in a directory was another directory containing entries then the parent list would get freed too early, causing a double free error or segfault.
Fix by ensuring that the completed list is at the top of the stack before freeing it. This will defer freeing parent lists until the contents of paths have been processed.
Lifecycle policies can cause the archive.info file and its copy to be removed since they are only updated on a stanza-upgrade. Update the timestamps after a successful backup to prevent this.
This does not mean that lifecycle policies should be used as a replacement for expiration. However, in some cases there may be policies in place that are out of admin control. If the lifecycle expiration is less than pgbackrest expiration then corruption of the earliest backup will occur at the very least and there might be other corruption which would make the repo unrecoverable.
An error that gets raised all the way to the top TRY block might need to free a lot of resources and any of these callbacks could throw an error and mask the original error. In fact this is pretty likely since we are already in an error state. For example, the Db object will try to close the remote db connection, but if the protocol is in a bad state it will not be able to do so.
Solve this, for now, by not freeing memory or calling callbacks in the CATCH_FATAL() block. This gives us a better chance if being able to report the error without encountering another error first.
For the most part, we don't need to worry about freeing resources (file handles, TLS contexts, etc.) if the program is going to exit immediately. However, it is important to attempt to terminate all active protocol connections, which is done by protocolFree() in main() since the protocol objects live in the top context.
Another way to handle this would be to implement an error stack and that is probably something we will do in the future. But, in the case of a segfault the original error would still be lost. Yet another option would be to still do cleanup but defer it until after the CATCH_FATAL() block.
If a repo is not specified for the expire command then the lowest repo becomes the default. The repo-retention-full value for time was being retrieved from the default rather than a specific repo which led to an incorrect expiration being applied.
Get the value from the specific repo and add a test.
It would be better if the default repo could not be queried in this case but it is not clear how to do that since the repo option is valid for expire (unlike, e.g., archive-push).
Allow key/value annotations to be added with the backup command and added/modified/removed with the new annotate command.
Annotations can be viewed with the info command in text mode when --set is specified and are always included in JSON output.
There are performance benefits to increasing the upload chunk size as long as the tradeoff with additional memory consumption is acceptable.
Make the chunk size configurable for S3, GCS, and Azure, but don't attempt to do any validation of the chunk size beyond some sane limits. The defaults remain as is for each storage type to avoid any unintentional regressions.
Catching individual fatal errors was only used in testing so the tests have been updated to use other errors instead. CATCH_FATAL() is now the only way to catch fatal errors.
This simplifies the logic a bit for upcoming changes to error handling and cleanup.
Also fix an issue where passing errorMessage() directly to THROW*() would attempt to copy the message buffer instead of preserving it, which is undefined behavior. Since there were no instances of this behavior before this commit, this was not a live bug.
All unit and performance tests are now built by the C harness.
Remove all unit/performance test build code from Perl.
Remove code from C harness that is no longer used. This code was included so the C harness could be run separately, but that is no longer needed with this full integration.
The C test harness is used for unit tests from the Perl harness where possible. Currently, unit tests can be run in the C harness when --no-coverage is specified and --profile is not specified.
C harness tests work on meson 0.45.
The C harness runs with valgrind by default. Valgrind can be disabled with --no-valgrind.
Also rebuild containers to add meson and update the documentation so that meson builds will work (even though we don't do them yet).
The standby timeline check was being performed using pg_control data loaded before the backup started. If the backup was started immediately after a promotion the standby might not have executed a checkpoint and written the new timeline to pg_control.
Instead perform the timeline check after the checkpoint is executed. This should ensure that the new timeline is in pg_control.
The prior warning made it sound as if some action was required on the part of the user.
The new message should make it clearer that this action will be performed by pgBackRest.
Build pgbackrest binary and auto-generated code automatically.
Remove --module option and allow modules to run by parameter. This is less verbose and multiple modules can be run at a time.
Allow filtering of modules. Multiple tests can be passed as parameters and if the module ends in / it will be used as a prefix filter. For example, common/ will run all the common modules.
If a test errors the remaining tests will still run but the test process will eventually exit with an error.
CI tests are included but unit tests remain on the development branch.
With these changes all unit tests run except those that specify the define (e.g. common/assert-off) or containerReq (e.g. protocol/protocol) keywords.
Building the C test harness has been simplified:
meson -Dwerror=true -Dfatal-errors=true -Dbuildtype=debug test/build/none pgbackrest
ninja -C test/build/none test/src/test-pgbackrest
To run all modules:
test/build/none/test/src/test-pgbackrest test
Just the common/error module:
test/build/none/test/src/test-pgbackrest test common/error
All info modules:
test/build/none/test/src/test-pgbackrest test info/
Add tzdata package so timezone tests in command/restore work correctly.
Mark default git path as safe. This is a security fix that is not applicable in this environment, but must be set.
Also remove package cleanup, which is inconvenient when new packages need to be installed. It makes sense for containers that will be downloaded from Dockerhub but not so much for a locally-maintained container.
This was clearly an attempt to set the mode when creating a directory, but it never worked and instead created a "750" directory in the current working directory.
Detected when running in an environment where the current working directory was read-only.
Add harness depends when present.
Include libyaml in all test builds.
Fix mode on paths before trying to remove and set test path with mode 770 to match the Perl test harness.
With these changes all unit tests run except those that specify the define (e.g. common/assert-off), binReq (e.g. command/archive-get), or containerReq (e.g. protocol/protocol) keywords.
Builds and code generation need to be done in advance. The following commands are required for setup:
meson setup -Dwerror=true -Dfatal-errors=true -Dbuildtype=debug build pgbackrest
ninja -C build test/src/test-pgbackrest
build/src/build-code help pgbackrest
build/src/build-code postgres pgbackrest
Now tests can be run, e.g.:
build/test/src/test-pgbackrest --module=postgres/interface
Creating new binaries was convenient at first but has now become a maintenance issue.
Solve this by combining that into a single binary that takes an additional parameter to indicate which code should be built.
Also clean up path handling to make it easier to build code from the command line.
This makes the test code a bit simpler where we are listing a path but not following links.
Links in the repository can be used for testing but should never be committed to the main branch.
NOTE TO PACKAGERS: An experimental meson build has been added but packagers should continue to use the autoconf/make build for the foreseeable future.
Improvements:
* OpenSSL 3 support. (Reviewed by Stephen Frost.)
* Create snapshot when listing contents of a path. (Reviewed by John Morris, Stephen Frost.)
* Force target-timeline=current when restore type=immediate. (Reviewed by Stephen Frost.)
* Truncate files during delta restore when they are larger than expected. (Reviewed by Stephen Frost.)
* Disable incremental manifest save when resume=n. (Contributed by Reid Thompson. Reviewed by David Steele.)
* Set backup percent complete to zero before copy start. (Contributed by Reid Thompson. Reviewed by David Steele.)
* Use S3 IsTruncated flag to determine list continuation. (Reviewed by John Morris, Soulou. Suggested by Christian Montagne.)
Documentation Bug Fixes:
* Skip internal options in the configuration reference. (Reported by Francisco Miguel Biete.)
Documentation Improvements:
* Add link to PostgreSQL configuration in repository host section. (Reviewed by Stefan Fercot. Suggested by Julien Cigar.)
Test Suite Improvements:
* Add experimental Meson build. (Reviewed by Eli Schwartz, Sam Bassaly.)
* Allow any path to be passed to the --test-path option. (Contributed by Andrey Sokolov. Reviewed by David Steele.)
* Fix compile error when DEBUG_EXEC_TIME is defined without DEBUG. (Contributed by Andrey Sokolov. Reviewed by David Steele.)
Explicitly set target timeline to "current" when type=immediate and PostgreSQL >= 12. We do this because type=immediate means there won't be any actual attempt to change timelines, but if we leave the target timeline as the default of "latest" then PostgreSQL might fail to restore because it can't reach the "latest" timeline in the repository from this backup.
This is really a PostgreSQL bug and will hopefully be addressed there, but we'll handle it here for older versions, at least until they aren't really seen in the wild any longer.
PostgreSQL < 12 defaults to "current" (but does not accept "current" as a parameter) so no need set it explicitly.
Previously a callback was used to list path contents and if no sort was specified then a snapshot was not required. When deleting files from the path some filesystems could omit files that still existed, which meant the path could not be removed.
Filter . out of lists in the Posix driver since this special entry was only used by test code (and filtered everywhere in the core code).
Also remove callbacks from the storage interface and replace with an iterator that should be easier to use and guarantees efficient use of the snapshots.
v0.45 ships with Ubuntu 18.04, which is currently the oldest distro we support. We may never do a Meson release on Ubuntu 18.04 but this allows us to start running unit tests with Meson in the meantime.
Some more granular options are not available so we use buildtype in more places.
The check for a in-tree autoconf/make build had to be removed since the filesystem APIs are not available.
Finally, alias_target was removed. This means that full paths must be used for build targets, which does not seem too bad. For instance, test/src/test-pgbackrest must now be used as a build target instead of simple test-pgbackrest.
Coverage for these checks was dependent on the order the files were read from disk, which made the tests fragile.
Rearrange the checks and add a test that won't depend on order.
Previously we were just checking for the existence of NextContinuationToken, which the S3 documentation indicates will not be present when the list is not truncated. However, recent versions of Scality send a blank NextContinuationToken when IsTruncated is false. Sending the blank continuation token back causes Scality to send another blank continuation token and an infinite loop occurs.
Instead use IsTruncated (which is required to be present) to determine whether NextContinuationToken should be present. Error if NextContinuationToken is then missing or empty, since an empty token caused an infinite loop with the Scality server (which arguably should have errored when passed an empty token).
The TEST_STORAGE_LIST() macro is more robust and hides the callback mechanism from the caller.
Add features to TEST_STORAGE_LIST() that hrnStorageInfoListCallback() had.
Update tests to use the abbreviated type output (e.g. path/) generated by TEST_STORAGE_LIST().
Having the test harness in C will allow us to remove duplicated Perl code and test on systems where Perl support is not present.
Custom harnesses and shims are currently not implemented, which means only the following tests in the common module will run: error, stack-trace, type-convert, assert-on, mem-context, time, encode, type-object, type-string, type-list, type-buffer, type-variant, reg-exp, log.
The experimental test harness is being committed with partial functionality so it can be used in Windows development. The remaining features will follow as needed.
The meson builds are still experimental so for now the configure/make build process is preferred for release builds. This message should help prevent any automated build systems from picking up meson instead.
Some of the replacements that were being done already existed as constants, so use the constants instead.
Also fix a minor formatting error introduced when testAdd() was renamed to hrnAdd().
This module has dependencies on command/command so it does not make sense for it to be in the common module. Also move protocolFree() to main() since this is a very large dependency.
Adjust the tests so command/exit can be tested later. This is a bit messy but will get adjusted as we improve the test harness.
Both have newer gcc and OpenSSL 3.
Fedora 36 runs horribly slow with valgrind enabled so run the valgrind tests on Ubuntu 22.04. Fedora 36 has a newer gcc so it is still worth testing on.
There are two changes:
* Suppress deprecation warnings so we can build with -Werror and -Wfatal-errors. At some point we'll need to migrate to the new APIs but there does not seem to be a good reason to support two sets of code right now.
* Update the handling for unexpected EOF to handle EOF or error. The error code for EOF has changed and become harder to identify, but we probably don't care whether it is an error or EOF.
Maintaining the version interfaces was complicated by the fact that each interface needed to be in separate compilation unit to avoid type conflicts. This also meant that various build/test files needed to be updated to add the new interfaces.
Solve these problems by auto-generating all the interfaces into a single file. This is made possible by parsing defines and types out of the header files and creating macros to rename the types. At the end of the version interface everything is undef'd. Another benefit is that the auto-generated interfaces can be static and included directly into postgres/interface.c.
Since some code generation is now always required for tests, change --no-gen to --min-gen in test.pl.
It would also make sense to auto-generate the version defines in postgres/version.h, but that will be left for a future commit.
Meson is a new build system that offers simpler syntax and superior performance to autoconf/make. In addition, Windows is supported natively.
The Meson build appears complete, but currently is used only for auto-generation of code and the host build of pgbackrest. Some container upgrades will be required before Meson can be used for container builds.
Also patch the Debian package to force autoconf/make rather than Meson.
Stopping the cluster has started consistently running out of memory on PostgreSQL 9.1. This seems to have happened after pulling in new packages at some point so it might be build related.
Stopping the cluster is not critical for 9.1 so skip it.
These files were never intended to be compiled on their own so the .c extension was a bit misleading. In particular Meson does not like .c files that are not intended to be compiled independently.
Leave header files as is since they are already protected against being included more than once and are never expected to be compiled.
The manifest is saved on a regular basis during a backup so a failed backup can be resumed. For backups that the user has configured/invoked as not resumable, skip the incremental save of the manifest.
Previously the behavior was to download the file from the repository when it was not exactly the same size in PGDATA. However, it may just be that the file was extended and the contents are the same up to the file size recorded in the manifest. This could also be very valuable for files that are always append only, like logs.
Change info.size to file->size in one place. Both are technically correct but file->size makes more sense.
Use the new fileName variable in a few existing places.
Also adjust some existing comments to make them clearer.
Remove VM_OS_REPO since it is no longer required.
Rebalance PostgreSQL versions for more efficient test times.
Always print version of PostgreSQL when testing. This helps verify that new minor releases are being used.
Each mem context can track child contexts, allocations, and a callback. Before this change memory was allocated for tracking all three even if they were not used for a particular context. This made mem contexts unsuitable for String and Variant objects since they are plentiful and need to be as small as possible.
This change allows mem contexts to be configured to track any combination of child contexts, allocations, and a callback. In addition, the mem context can be configured to track a single child context and/or allocation, which saves memory and is a common use case.
Another benefit is that Variants can own objects (e.g. KeyValue) that they encapsulate. All of this makes memory accounting simpler because mem contexts have names while allocations do not. No more memory is used than before since Variants and Strings still had to store the memory context they were originally allocated in so they could be easily freed.
Update the String and Variant objects to use this new functionality. The custom strFree() and varFree() functions are no longer required and can now be a wrapper around objFree().
Lastly, this will allow strMove() and varMove() to be implemented and used in cases where strDup() and varDup() are being used to move a String or Variant to a new context. Since this will be a bit noisy it is saved for a future commit.
Because there is a lot of repetition in this file, changes can look very jumbled with existing data in a diff. Also, if can be hard to tell what is being modified if the diff does not show enough lines before and after.
This change adds labels to the end of the line to localize the diff and make it easier to see what has been changed. Also, remove some linefeeds and make separators more consistent.
The change to parse.auto.c will be committed separately so it can be ignored in history/blame.
Bug Fixes:
* Fix error thrown from FINALLY() causing an infinite loop. (Reviewed by Stephen Frost.)
* Error on all lock failures except another process holding the lock. (Reviewed by Reid Thompson, Geir Råness. Reported by Geir Råness.)
Features:
* Backup file bundling for improved small file support. (Reviewed by Reid Thompson, Stefan Fercot, Chris Bandy.)
* Verify command to validate the contents of a repository. (Contributed by Cynthia Shang, Reid Thompson. Reviewed by David Steele, Stefan Fercot.)
* PostgreSQL 15 support. (Reviewed by Stefan Fercot.)
* Show backup percent complete in info output. (Contributed by Reid Thompson. Reviewed by David Steele.)
* Auto-select backup for restore command --type=lsn. (Contributed by Reid Thompson. Reviewed by Stefan Fercot, David Steele.)
* Suppress existing WAL warning when archive-mode-check is disabled. (Contributed by Reid Thompson. Reviewed by David Steele.)
* Add AWS IMDSv2 support. (Contributed by Nuno Pires. Reviewed by David Steele.)
Improvements:
* Allow repo-hardlink option to be changed after full backup. (Reviewed by Reid Thompson.)
* Increase precision of percent complete logging for backup and restore. (Contributed by Reid Thompson. Reviewed by David Steele.)
* Improve path validation for repo-* commands. (Contributed by Reid Thompson. Reviewed by David Steele.)
* Improve stop command to honor stanza option. (Contributed by Reid Thompson. Reviewed by David Steele. Suggested by ragaoua.)
* Improve error message for invalid repo-azure-key. (Contributed by Reid Thompson. Reviewed by David Steele. Suggested by Seth Daniel.)
* Add hint to check the log on archive-get/archive-push async error. (Reviewed by Reid Thompson.)
* Add ClockError for unexpected clock skew and timezone changes. (Reviewed by Greg Sabino Mullane, Stefan Fercot. Suggested by Greg Sabino Mullane.)
* Strip extensions from history manifest before showing in error message. (Reviewed by Stefan Fercot.)
* Add user:group to lock permission error. (Reviewed by Reid Thompson.)
Documentation Bug Fixes:
* Fix incorrect reference to stanza-update in the user guide. (Fixed by Abubakar Mohammed. Reviewed by David Steele.)
* Fix example for repo-gcs-key-type option in configuration reference. (Reviewed by Reid Thompson.)
* Fix tls-server-auth example and add clarifications. (Reviewed by Reid Thompson.)
Documentation Improvements:
* Simplify messaging around supported versions in the documentation. (Reviewed by Stefan Fercot, Reid Thompson, Greg Sabino Mullane.)
* Add option type descriptions. (Contributed by Reid Thompson. Reviewed by David Steele.)
* Add FAQ about backup types and restore speed. (Contributed by David Christensen. Reviewed by Reid Thompson.)
* Document required base branch for pull requests. (Contributed by David Christensen. Reviewed by Reid Thompson.)
If the user requested the exact repo path then strSub() would be passed an invalid start value leading to an assertion:
$ pgbackrest --stanza=test repo-ls /var/lib/pgbackrest
ASSERT: [025]: start <= this->pub.size (on dev builds)
ASSERT: [025]: string size must be <= 1073741824 bytes (on prod builds)
Fix this by checking if the requested path exactly equals the repo path and returning an empty relative path in this case.
Another issue was that invalid subpaths were not detected if they started with the repo path. For example, /var/lib/pgbackrestsub would not generate an error if the repo path was /var/lib/pgbackrest. Fix this by explictly checking for a / between the repo path and the subpath. This also requires special handling when the repo path is /.
This is not a live bug since the issues were found in an unreleased feature introduced in 5ae84d5.
The encrypted archive-push and repo tests were running very slowly on 32-bit with Valgrind enabled. This appears to be an issue with a newer version of Valgrind, but it has been going on long enough that bisecting does not seem to be worthwhile.
Reduce the size of the encrypted test segments where possible to improve overall test performance.
Integration expect log testing was originally used as a rough-and-ready way to make sure that certain code paths were being executed before the unit tests existed. Now that we have 100% unit test coverage (with expect log testing) the value of the integration expect tests seems minimal at best.
But they do cause numerous issues:
- Maintenance of the expect code and replacements that are required to keep logs reproducible.
- Even a trivial change can cause massive churn in the expect logs, e.g. d9088b2. These changes should be minutely audited but since the expect logs have little value now it is seldom worth the effort.
- The OS version used to do expect testing (RHEL7) can only be used to test one version of PostgreSQL. This makes it hard to balance the PostgreSQL version testing between OS versions.
- When a commit affects expect logs it is not clear (especially for new developers) how to regenerate them and our contributing guide is silent on the issue.
The goal is to migrate the integration tests to C and expect testing is not part of that plan. It seems best to get rid of them now.
Once upon a time the allocation array was allocated up front so this test was required for the top context, which did not allocate up front.
Now allocations are done on demand so this case is covered for every context that does not allocate memory.
This helps rebalance some of the tests that are running long, i.e. d9 and u20.
I would be better to move more PostgreSQL versions to d9, but the base VM does not contain more versions. New minor versions will be out later in the week so that seems a better time to be rebuilding containers.
The emulation is so slow that running all the unit tests would be too expensive, but this at least shows that the build works and some of the more complex tests run. In particular, it is good to test on one big-endian architecture to be sure that checksums are correct.
Update checksums in the tests where they had gotten out of date since the last time we were testing on s390x. Also use a different test in command/archivePushTest to show the name of the file when a checksum does not match to aid in debugging.
The command/archive-push test was updated but not included because there is also a permissions issue, which looks to be the same as what we see on MacOS/FreedBSD. Hopefully we'll be able to fix all of those at the same time.
The function worked fine, but Coverity was unable to determine that the finally block was run, which led to false positives about unfreed memory.
Using a boolean in the block makes it clear to Coverity that the finally block will always be run no matter what else happens.
We'll depend on the compiler to optimize away the boolean if it is not used in a finally block. The cost of the boolean is fairly low in comparison to everything else being done in these macros, so it does not seem worth having a separate block even if the compiler is not able to eliminate the boolean.
This reverts most of 9a271e9 that fixed a bug caused by c5b5b58, which was also attempting to help Coverity understand FINALLY() blocks.
