This repository has been archived by the owner on Dec 16, 2022. It is now read-only.
slack vitess update 10.02.2017 #40
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The recent v3 scatter aggregates feature introduced a regression: If there is a complex aggregate expression, but the group by references a unique vindex, then it fails the query as unsupported. The fix was tricky because the newer code creates the OrderedAggregate primitive upfront, and then it gets dissolved later if a group by was found referencing a unique vindex. But this was possible only for non-complex expressions where the OA primitive could represent them. The newer code instead looks ahead into the group by first. If a unique vindex is found, then no OA is created. So, complex expressions can be freely pushed down into the route. This complicates the look-ahead code because it has to do it without disturbing the symbol table. But it also simplifies the OA code a bit because there's no need to dissolve it in the future. The execution tree is essentially fixed upfront.
…-reset-shard-targetting Reset shard targeting using USE with no argument
…meout makes the various timeouts configurable on the connection -- connect,…
bug: reset all sequences when mastership is lost
…ntifiers When tablet state changes, the logs currently include a verbose dump of the Tablet as well as the before and after Target objects. Use the existing TabletIdent and add an analogous TargetIdent helper function to print more concise versions of the same information.
The error message added in b5339b7 didn't include format string entries for all the arguments, so add them for a nicer warning message.
don't pick up python2 from PYTHONPATH during selenium install
two minor logging changes
v3: fix regression in aggregate handling
* Change ExprFromValue to use categorization functions. * Added unit test for ExprFromValue. * Fixed bit default value handling, with unit and endtoend tests.
fix typo in acl comment.
Couple fixes
It is too flaky at this point because tests are running too slow and we run into test timeouts hardcoded in the unit test. Before we try to re-enable this, let me switch the Travis tests to run in the new VM infrastructure via Docker first.
vttablet: Hot Row Protection: Make the transaction concurrency for a hot row configurable.
travis: Temporarily disable the unit_race test on Travis.
throttler: client unit test: Register the test server before calling gRPC's Serve().
BUG=67060474 The following deadlock is possible: 1. runSend obtains a lock. 2. Poller gets trigerred, and waits for lock. 3. runSend calls this poller trigger, but the trigger will hang because it cannot return until poller returns. Unfortunately, this race is not reproduceable in a test because it's a CPU vs. CPU race. So, we just have to rely on reasoning.
Change the test options to enable normalization by default. Fix typos in test case strings to use ' not " Add a couple more tests
messages: fix deadlock
fix vtexplain tests
Based on real-life workloads, we found that it may not be a good idea to dedup all values. Specifically, values used in DMLs like INSERT etc. should not be deduped. They actually end up polluting the plan cache with all kinds of combinations. With the change, values are deduped only if they are within selects. This deduping happens even if there are subqueries withing DMLS, while the DML parts are still not deduped. Additionally, it doesn't make sense to take the effort to dedup values that are too long. So, I've added a check where if a value is longer than 256 bytes, we blindly create a new bind var.
…ip-invalid-keyspace make show vitess_shards not fail if one keyspace is invalid
v3: normalizer: smarter dedup
index fields by name, original name, and full name
…-vtgaate Added query cache metrics
…ier-message ParseTabletAlias: Help the user by explaining the expected format.
…r-query-plan-cache Adds session option to cache query plans
vttablet's ActionAgent has three startup methods: 1. real production tablet 2. vtcombo in-memory tablet 3. in-memory test tablet for unit tests The optional RebuildKeyspace when SrvKeyspace doesn't exist only needs to be done in case .1, not 2. and 3. vtcombo does an explicit rebuild after having fully created the keyspace (so we don't need the background intermediate ones that will fail and spam the logs anyway). in-memory tests don't ususally need the SrvKeyspace (and if they do, they could call it).
Moving optional RebuildKeyspace out of tests.
|
LGTM |
demmer
force-pushed
the
slack-vitess-10.02.2017
branch
from
360abe9 to
1dec806
Compare
|
oh - it looks like we forgot to cherry pick this: 360abe9 |
|
I updated the PR description accordingly. |
|
👍 |
rafael
pushed a commit
that referenced
this pull request
Dec 11, 2019
Signed-off-by: Ajeet jain <[email protected]> * readme for go endtoend test cases Signed-off-by: Ajeet jain <[email protected]> * Update README.md Signed-off-by: Arindam Nayak <[email protected]>
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Branch to sync with upstream and then apply three slack-specific patches onto the
slack-new-masterbranch:d09dcd9 redact out the /txlogz endpoint completely
9114772 redact out the /debug/consolidations UI
1dec806 redact bind variables and the full sql from /debug/querylog
Once this is merged I will force push to master.