Граф коммитов

9747 Коммитов

Автор SHA1 Сообщение Дата
Michael Berlin 8980006d1b discovery: Fix datarace for "listener" field in HealthCheckImpl.
The fix is to treat "listener" as an immutable field which must not be changed once initially written.
2016-05-23 14:24:03 -07:00
Alain Jobart 900ba27ab1 Folding actionnode into vt/topo. 2016-05-23 14:06:34 -07:00
Alain Jobart cd852e9215 Renaming ActionNode to Lock so it can move. 2016-05-23 13:19:35 -07:00
Alain Jobart 0b49bfe44a Refactoring ActionNode, before move.
Spliting up the TabletAction constants into tabletmanager, where they
more naturally belong. That removes a bunch of dependencies.

Now actionnode is only used for locking and unlocking shards and
keyspaces. Will move it to a better place next.
2016-05-23 10:19:14 -07:00
Alain Jobart cdf354b8c3 Removing useless actionnode.SlaveWasRestartedArgs.
It was the only place we used a struct for parameter passing.
And there is only one parameter. Go figure.
2016-05-23 09:38:17 -07:00
Alain Jobart 60b60837c1 Re-working locking for topo server.
Now using context methods to keep tarck of the locks, and allow checking
for them.
Documenting what we lock for in the new methods.
Dropping shard locks for a lot of operations, as the keyspace lock is
enough.
2016-05-23 09:15:29 -07:00
Alain Jobart 536f6a48a4 Now locking keyspace for MigrateServedTypes. 2016-05-23 09:15:29 -07:00
Anthony Yeh 9b93e53741 Merge pull request #1712 from enisoc/mysql57
Support MySQL 5.7
2016-05-20 22:56:53 -07:00
sougou 4af82f384e Merge pull request #1720 from sougou/bugs
bugs: b/28851938, b/28888701
2016-05-20 18:49:14 -07:00
Sugu Sougoumarane 2219353220 bugs: 30->32 2016-05-20 18:34:14 -07:00
Sugu Sougoumarane f5f689e802 bugs: b/28851938, b/28888701
The mysql API bug seems to trigger for very large data sizes,
likely greater than 1<<20. This is because we make go think
that those slices can't be bigger than that value.

The other fix is to prevent a divide by 0 error.
2016-05-20 18:21:27 -07:00
Anthony Yeh 12f29bc5a1 Enable relay_log_recovery for MySQL 5.6+.
This causes mysqld to ignore relay logs on disk at startup,
and instead start replicating from the master wherever the SQL thread
left off.

This is generally recommended for safer crash recovery.
We also do it to avoid problems caused by spurious Previous-GTIDs lists
in relay logs that occur when we restart mysqld in integration tests.
2016-05-20 17:59:37 -07:00
Anthony Yeh ce361a9ccc Protect new system tables from deletion in test cleanup. 2016-05-20 17:59:37 -07:00
Anthony Yeh fd5cac0879 For MySQL 5.7, use 'mysqld --initialize' instead of mysql_install_db.
The latter is deprecated. Also check for both pid and sock files to
disappear when shutting down, or else there is a race.
2016-05-20 17:59:37 -07:00
Anthony Yeh 3d22fa35ee go/mysql: Support MySQL 5.7 2016-05-20 17:59:37 -07:00
Anthony Yeh b230519386 Remove my.cnf options that were deprecated in MySQL 5.6.
These options were removed entirely in MySQL 5.7.
2016-05-20 17:59:37 -07:00
Anthony Yeh 4170be37ce Add Docker images for MySQL 5.7. 2016-05-20 17:59:37 -07:00
Anthony Yeh dea4eb2929 Allow MySQL56 flavor to accept MySQL 5.7. 2016-05-20 17:59:37 -07:00
Anthony Yeh f8ba1f84a5 Document how to build custom Docker images for Kubernetes. 2016-05-20 15:27:43 -07:00
Anthony Yeh 2b363f13d5 Merge pull request #1718 from enisoc/mysqlctl-wait-time
Increase default mysqlctl wait_time to 5min.
2016-05-20 14:09:26 -07:00
Anthony Yeh 9ace4b7106 mysqlctld: Exit daemon if mysqld init/start fails.
Also make sure to log the reason it failed.
2016-05-20 13:44:34 -07:00
sougou 90b64810fb Merge pull request #1717 from sougou/export_internal_changes
Exporting internal changes back to open-source.
2016-05-20 12:03:54 -07:00
Anthony Yeh 6d7e41909d Increase default mysqlctl wait_time to 5min.
Especially on test setups, mysqld can take a while to initialize.
We'd rather wait a little longer to find out if it's truly stuck,
rather than give up too soon when it's actually fine but slow.
2016-05-20 11:58:29 -07:00
Sugu Sougoumarane 6e896cf69b Exporting internal changes back to open-source.
NOTE: This is an automated export. Changes were already LGTM'd internally.
2016-05-20 11:46:46 -07:00
Michael Berlin 1f32212c8e Merge pull request #1707 from michael-berlin/throttler_filtered_replication_plumbing
binlogplayer: Enable throttler in filtered replication.
2016-05-20 11:10:33 -07:00
sougou 79aaac1af7 Merge pull request #1708 from sougou/v3
v3: VSchemaFormal -> proto3
2016-05-20 10:27:45 -07:00
Michael Berlin 35831e3320 Merge pull request #1715 from michael-berlin/worker_test_assertIn_fix
worker: Fix assertIn check (parameters were swapped).
2016-05-20 10:27:34 -07:00
Michael Berlin 6f9d6f1780 tabletmanager: Fixed missing mutex unlock in BinlogPlayerController.Close(). 2016-05-20 10:04:51 -07:00
Sugu Sougoumarane 78fcdf8f20 Merge branch 'master' into v3 2016-05-20 10:00:01 -07:00
Sugu Sougoumarane 1b1ec26f2f v3 vschema: typo in comment 2016-05-20 09:56:05 -07:00
Alain Jobart 923ca110a8 Merge pull request #1711 from alainjobart/splitshardcount
Splitshardcount
2016-05-20 09:37:20 -07:00
Alain Jobart 2b8b9d06c3 Merge pull request #1713 from alainjobart/rep
Adding doc for replication.
2016-05-20 08:41:15 -07:00
Alain Jobart ed0013a3af Integrating comments from Anthony. 2016-05-20 08:40:42 -07:00
Sugu Sougoumarane 74295d38a0 v3 vschema: "Sequence" -> "sequence" 2016-05-19 23:55:44 -07:00
Sugu Sougoumarane dd30f0b5a1 v3 vschema: code review comments 2016-05-19 23:34:55 -07:00
Michael Berlin 7405126abb binlog: Do not duplicate error strings in the code. 2016-05-19 16:29:13 -07:00
Michael Berlin 476d097f44 worker: Fix assertIn check (parameters were swapped).
Before this, the test checked if the empty string (stdout) is in the (actually) wanted text. This always succeeded even if it should not have.
2016-05-19 15:12:16 -07:00
Michael Berlin 57ce093076 Merge pull request #1714 from michael-berlin/export_histograms_upper_bounds
Exporting internal changes back to open-source.
2016-05-19 15:57:16 -05:00
Michael Berlin e3524b301c Exporting internal changes back to open-source.
NOTE: This is an automated export. Changes were already LGTM'd internally.
2016-05-19 13:38:53 -07:00
Michael Berlin b6076a803d tabletmanager: Change behavior for Close() on non-stopped BinlogPlayerController from killing the process to returning an error.
This is more in line with the rest of the code which also catches panics and logs them instead of crashing the whole process.
2016-05-19 13:30:28 -07:00
Michael Berlin e6e64d062d binlog: Clarified why we don't check for context cancellation while being throttled. 2016-05-19 13:30:28 -07:00
Michael Berlin 0cc07c32da throttler: Remove explicit cast from constants because it's not necessary. 2016-05-19 13:30:28 -07:00
Michael Berlin 1f82edf7ee throttler: Clarify that the maximum backoff duration is 1 second. 2016-05-19 13:30:17 -07:00
Alain Jobart af0db4044e Adding doc for replication. 2016-05-19 13:08:09 -07:00
Anthony Yeh 9194165d5f k8s: Use 'DeleteShard -recursive' instead of individual DeleteTablets.
The individual DeleteTablet calls in vttablet-down.sh were from before
we had the -recursive flag on DeleteShard, which now does that for you.
2016-05-19 12:12:26 -07:00
Anthony Yeh d776a363a4 doc: Don't use SetKeyspaceShardingInfo in V3 resharding. 2016-05-19 12:04:48 -07:00
Alain Jobart 8708e91530 Using FindAllShardsInKeyspace to fix k8 env.
Also adding an integration test for it, just to be sure.
2016-05-19 12:01:59 -07:00
Alain Jobart d73c337514 Removing keyspace.split_shard_count. Unused. 2016-05-19 09:36:06 -07:00
Alain Jobart 07b0a78823 Small fixes in servenv. 2016-05-19 08:08:42 -07:00
Alain Jobart cc6e894625 Merge pull request #1706 from alainjobart/srvshard
Removing SrvShard
2016-05-19 07:38:11 -07:00