This patch mainly do the following refactors:
- Use int64_t instead of int32_t for 'version' in DeleteHandler
- Move some comments from .cpp to .h file, add some new comments in .h files, and also remove some meaningless comments
- Use switch...case... instead of multiple if..else.. for DeleteConditionHandler::is_condition_value_valid
- Use range loop to simplify code
- Reduce some compare operations in Cond::del_eval
- Improve some branch predictions in Reader
- Fix and improve some unit tests
The current compaction mechanism is that there is a producer thread that has been producing compaction tasks,
and the selected tablet must apply for `permits`.
When a tablet could hold `permits`, compaction task for this tablet will be submitted to thread pool.
We take compaction score as `permits` which is used for limiting memory consumption.
However, `pick_rowset_to_compaction()` will be executed before the file merge in compaction thread,
and the number of segment files that actually perform the merge operation is smaller than compaction score.
In addition, it is also possible that compaction task exits directly because the tablet doesn't meet
the requirements of compaction.
This patch optimizes and refactors the code of compaction, so that we can execute 'pick rowsets'
before applying for permits for a compaction task, calculate the number of segment files that actually
participate in the merge operation, and take this number as `permits`.
A large number of small segment files will lead to low efficiency for scan operations.
Multiple small files can be merged into a large file by compaction operation.
So we could take the tablet scan frequency into consideration when selecting an tablet for compaction
and preferentially do compaction for those tablets which are scanned frequently during a
latest period of time at the present.
Using the compaction strategy of Kudu for reference, scan frequency can be calculated
for tablet during a latest period of time and be taken into consideration when calculating compaction score.
Currently, there are M threads to do base compaction and N threads to do cumulative compaction for each disk.
Too many compaction tasks may run out of memory, so the max concurrency of running compaction tasks
is limited by semaphore.
If the running threads cost too much memory, we can't defense it. In addition, reducing concurrency to avoid OOM
will lead to some compaction tasks can't be executed in time and we may encounter more heavy compaction.
Therefore, concurrency limitation is not enough.
The strategy proposed in #3624 may be effective to solve the OOM.
A CompactionPermitLimiter is used for compaction limitation, and use single-producer/multi-consumer model.
Producer will try to generate compaction tasks and acquire `permits` for each task.
The compaction task which can hold `permits` will be executed in thread pool and each finished task will
release its `permits`.
`permits` should be applied for before a compaction task can execute. When the sum of `permits`
held by executing compaction tasks reaches a threshold, subsequent compaction task will be no longer allowed,
until some `permits` are released. Tablet compaction score is used as `permits` of compaction task here.
To some extent, memory consumption can be limited by setting appropriate `permits` threshold.
Fix stale path delete checking logic.
When current main path is version missing, then delete checking logic is always core dumped. So we fix the checking logic to tolerate current main version missing.
(1) fix recover persistent stale rowsets bug from multi-single version rowset in stale rowsets
(2) delete_expired_inc_rowsets check consistent version convert to [0, max_version]
Sometimes we want to detect the hotspot of a cluster, for example, hot scanned tablet, hot wrote tablet,
but we have no insight about tablets in the cluster.
This patch introduce tablet level metrics to help to achieve this object, now support 4 metrics on tablets: `query_scan_bytes `, `query_scan_rows `, `flush_bytes `, `flush_count `.
However, one BE may holds hundreds of thousands of tablets, so I add a parameter for the metrics HTTP request,
and not return tablet level metrics by default.
Persistence stale rowsets meta. When BE reboots, stale rowsets meta
can resume and the stale version can also be readable before stale gc time.
ISSUE: #4453
Compaction rules optimization, the detail problem description and design to see #4164.
This pr commits 2 functions:
(1) add the cumulative policy configable, and implement original policy.
(2) implement universal policy, the optimization version in #4164.
Sometimes we want to detect the hotspot of a cluster, for example, hot scanned tablet, hot wrote tablet,
but we have no insight about tablets in the cluster.
This patch introduce tablet level metrics to help to achieve this object, now support 4 metrics on tablets: `query_scan_bytes `, `query_scan_rows `, `flush_bytes `, `flush_count `.
However, one BE may holds hundreds of thousands of tablets, so I add a parameter for the metrics HTTP request,
and not return tablet level metrics by default.
In some very special circumstances, such as code bugs, or human misoperation, etc.,
all replicas of some tablets may be lost. In this case, the data has been substantially lost.
However, in some scenarios, the business still hopes to ensure that the query will not
report errors even if there is data loss, and reduce the perception of the user layer.
At this point, we can use the blank Tablet to fill the missing replica to ensure that the query can be executed normally.
Add a new FE config `recover_with_empty_tablet`. default is false. true means to use empty tablet to fill the missing one.
Also fix a bug in Fix#4274
Fix calculation of cumulative point. The problem is calculation of cumulative point
is wrong when be restarts and there is delete rowset. also see #4258
Related issue #4017, main changes as follows:
1. Add expired_snapshot_rs_version_map，_expired_snapshot_rs_metas，
2. Add VersionedRowsetTracker record compacted path version
3. Record path version when rowsets compact
4. In gc process, add expired snapshot rowsets to unused set to remove.
* [Bug] Fix bug that tablet meta lock twice
The tablet meta load may already be hold before calling
generate_tablet_meta_copy(), so we need provide a unlocked
version of generate_tablet_meta_copy()
* fix typo
Co-authored-by: chenmingyu <chenmingyu@baidu.com>
Current cumulative point calculate algorithm may skip singleton rowset when the rowset has only one segment and with NONOVERLAPPING flag. When a tablet is new created and cumulate many singleton rowsets, cumulative point will be calculated as the max version + 1, and then cumulative compaction couldn't pick any rowsets and compaction failed, and
will lead the next base compaction on this tablet with all rowsets, which can also cause memory consume problem, suppose there are thousands of rowsets.
All singleton rowsets must be newly wrote by delta writer and hasn't
do any compaction, we should place cumulative point before any of these rowsets.
* [Fix] Fix bug that rowset meta is deleted after compaction
After compaction, the tablet rowset meta will be modified by
adding to new output rowsets and deleting the old input rowsets.
The output version may equals to the input version.
So we should delete the "input" version from _rs_version_map
before adding the "output" version to _rs_version_map. Otherwise,
the new "output" version will be lost in _rs_version_map.
Adding a new storage engine, we need to make an extensible tablet interface, so olap/StorageEngine can support and manage new tablet types.
To start, this commit creates a class BaseTablet and make Tablet and new MemTablet inherit
this base class, some common fields & methods are moved to BaseTablet class, which fields
and methods belong to base/old class is not finalized yet, it will change as the project evolves.
Fix#3384
There is no functional changes in this patch.
Key refactor points are:
- Remove meaningless return value of functions in class Tablet, and
also some related functions in other classes
- Allow RowsetGraph::capture_consistent_versions to pass a nullptr
to the output parameter
- Use CHECK instead of LOG(FATAL) to simplify code
This CL fixes#3270 by skipping recently added version when performing cumulative compaction. A new config named "cumulative_compaction_skip_window_seconds" is added to adjust the time window.
When calculating the cumulative point at first time, we should stop increasing
the cumulative point when we meet a rowset with overlap flag as OVERLAPPING,
even if it has only one segments.
This CL try to fix a potential bug describe in ISSUE: #3097. But I'm not sure this is the root cause.
Also remove lots of verbose log, and fix a memory leak.
1. Add some comments to make the code easier to understand;
2. Make the metric `create_tablet_requests_failed` to be accurate;
3. Some internal methods use naked pointers directly instead of `shared_ptr`;
4. The `using` in `.h` files are contagious when included by other files,
so we should only use it in `.cpp` files;
5. Some formatting changes: such as wrapping lines that are too long
6. Parameters that need to be modified, use pointers instead of references
No functional changes in this patch.
When using an iterator of _tablet_map.tablet_arr(`std::list`) to remove
a tablet, we should first remove tablet from _partition_map to avoid
the iterator becoming invalid.
* Improve comparison and printing of Version
There are two members in `Version`:` first` and `second`.
There are many places where we need to print one `Version` object and
compare two `Version` objects, but in the current code, these two members
are accessed directly, which makes the code very tedious.
This patch mainly do:
1. Adds overloaded methods for `operator<<()` for `Version`, so
we can directly print a Version object;
2. Adds the `cantains()` method to determine whether it is an containment
relationship;
3. Uses `operator==()` to determine if two `Version` objects are equal.
Because there are too many places need to be modified, there are still some
naked codes left, which will be modified later.
This patch also removes some necessary header file references.
No functional changes in this patch.
time(NULL) returns second-resolution timestamp, however all compaction related time in Tablet are in millis-resolution. Therefore should use UnixMillis() instead.
Support compaction operation to compact only one rowset.
After the modification, the last rowset of the tablet will
also be compacted.
At the same time, we added a `segments_overlap_pb` field to
the rowset meta. Used to describe whether the segment data
in the rowset overlaps. This field is set by `rowset_writer`.
Initially UNKNOWN for compatibility with existing data.
In addition, the version hash of the rowset generated after
compaction is directly set to the version hash of last rowset
participating in compaction, to ensure that the tablet's
version hash remains unchanged after compaction.
When there are to many segment in one rowset, which is larger than
BE config 'max_cumulative_compaction_num_singleton_deltas', the
cumulative compaction will not work and just increase the cumulative
point, because there is only once rowset being selected.
So when selecting rowset for cumulative compaction, we should meet 2
requirments before finishing the selection logic:
1. compaction score is larger than 'max_cumulative_compaction_num_singleton_deltas'
2. at least 2 rowsets are selected.
The current compaction selection strategy and cumulative point update logic
will cause the cumulative compaction to not work, and all compaction tasks
will be completed only by the base compaction. This can cause a large number
of data versions to pile up.
In the current cumulative point update logic, when a cumulative cannot select
enough number of rowsets, it will directly increase the cumulative point.
Therefore, when the data version generates the same speed as the cumulative
compaction polling, it will cause the cumulative point to continuously increase
without triggering the cumulative compaction.
The new strategy mainly modifies the update logic of cumulative point to ensure
that the above problems do not occur. At the same time, the new strategy also
takes into account the problem that compaction cannot be performed if cumulative
points stagnate for a long time. Cumulative points will be forced to increase
through threshold settings to ensure that compaction has a chance to execute.
Also add a new HTTP API to view the compaction status of specified tablet.
See `compaction-action.md` for details.
Add a flag in RowsetMeta to record whether it has been deleted from rowset meta.
Before this PR, 37156 rowsets only cost 1642 s.
With this PR, 37319 rowsets just cost 1 s.
[Metric] Add tablet compaction score metrics
Backend:
Add metric "tablet_max_compaction_score" to monitor the current max compaction
score of tablets on this Backend. This metric will be updated each time
the compaction thread picking tablets to compact.
Frontend:
Add metric "tablet_max_compaction_score" for each Backend. These metrics will
be updated when backends report tablet.
And also add a calculated metric "max_tablet_compaction_core" to monitor the
max compaction core of tablets on all Backends.
