-
Notifications
You must be signed in to change notification settings - Fork 275
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
puller(ticdc): fix wrong update splitting behavior after table scheduling #11269
Conversation
Signed-off-by: dongmen <414110582@qq.com>
Signed-off-by: dongmen <414110582@qq.com>
Signed-off-by: dongmen <414110582@qq.com>
26b99b2
to
6a88c7c
Compare
/test dm-integration-test |
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: asddongmen, CharlesCheung96 The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
[LGTM Timeline notifier]Timeline:
|
/test dm-integration-test |
2 similar comments
/test dm-integration-test |
/test dm-integration-test |
/test cdc-integration-pulsar-test |
In response to a cherrypick label: base branch (release-7.5) needs to differ from target branch (release-7.5). |
Signed-off-by: ti-chi-bot <ti-community-prow-bot@tidb.io>
In response to a cherrypick label: new pull request created to branch |
In response to a cherrypick label: new pull request created to branch |
Signed-off-by: ti-chi-bot <ti-community-prow-bot@tidb.io>
In response to a cherrypick label: new pull request created to branch |
Signed-off-by: ti-chi-bot <ti-community-prow-bot@tidb.io>
commit c092599 Author: Ti Chi Robot <ti-community-prow-bot@tidb.io> Date: Wed Jun 12 00:26:59 2024 +0800 pkg/config, sink(ticdc): support output raw change event for mq and cloud storage sink (pingcap#11226) (pingcap#11290) close pingcap#11211 commit 3426e46 Author: Ti Chi Robot <ti-community-prow-bot@tidb.io> Date: Tue Jun 11 19:40:29 2024 +0800 puller(ticdc): fix wrong update splitting behavior after table scheduling (pingcap#11269) (pingcap#11282) close pingcap#11219 commit 2a28078 Author: Ti Chi Robot <ti-community-prow-bot@tidb.io> Date: Tue Jun 11 16:40:37 2024 +0800 mysql(ticdc): remove error filter when check isTiDB in backend init (pingcap#11214) (pingcap#11261) close pingcap#11213 commit 2425d54 Author: Ti Chi Robot <ti-community-prow-bot@tidb.io> Date: Tue Jun 11 16:40:30 2024 +0800 log(ticdc): Add more error query information to the returned error to facilitate users to know the cause of the failure (pingcap#10945) (pingcap#11257) close pingcap#11254 commit 053cdaf Author: Ti Chi Robot <ti-community-prow-bot@tidb.io> Date: Tue Jun 11 15:34:30 2024 +0800 cdc: log slow conflict detect every 60s (pingcap#11251) (pingcap#11287) close pingcap#11271 commit 327ba7b Author: Ti Chi Robot <ti-community-prow-bot@tidb.io> Date: Tue Jun 11 11:42:00 2024 +0800 redo(ticdc): return internal error in redo writer (pingcap#11011) (pingcap#11091) close pingcap#10124 commit d82ae89 Author: Ti Chi Robot <ti-community-prow-bot@tidb.io> Date: Mon Jun 10 22:28:29 2024 +0800 ddl_puller (ticdc): handle dorp pk/uk ddl correctly (pingcap#10965) (pingcap#10981) close pingcap#10890 commit f15bec9 Author: Ti Chi Robot <ti-community-prow-bot@tidb.io> Date: Fri Jun 7 16:16:28 2024 +0800 redo(ticdc): enable pprof and set memory limit for redo applier (pingcap#10904) (pingcap#10996) close pingcap#10900 commit ba50a0e Author: Ti Chi Robot <ti-community-prow-bot@tidb.io> Date: Wed Jun 5 19:58:26 2024 +0800 test(ticdc): enable sequence test (pingcap#11023) (pingcap#11037) close pingcap#11015 commit 94b9897 Author: Ti Chi Robot <ti-community-prow-bot@tidb.io> Date: Wed Jun 5 17:08:56 2024 +0800 mounter(ticdc): timezone fill default value should also consider tz. (pingcap#10932) (pingcap#10946) close pingcap#10931 commit a912d33 Author: Ti Chi Robot <ti-community-prow-bot@tidb.io> Date: Wed Jun 5 10:49:25 2024 +0800 mysql (ticdc): Improve the performance of the mysql sink by refining the transaction event batching logic (pingcap#10466) (pingcap#11242) close pingcap#11241 commit 6277d9a Author: dongmen <20351731+asddongmen@users.noreply.github.com> Date: Wed May 29 20:13:22 2024 +0800 kvClient (ticdc): revert e5999e3 to remove useless metrics (pingcap#11184) close pingcap#11073 commit 54e93ed Author: dongmen <20351731+asddongmen@users.noreply.github.com> Date: Wed May 29 17:43:22 2024 +0800 syncpoint (ticdc): make syncpoint support base64 encoded password (pingcap#11162) close pingcap#10516 commit 0ba9329 Author: Ti Chi Robot <ti-community-prow-bot@tidb.io> Date: Wed May 29 09:07:21 2024 +0800 (redo)ticdc: fix the event orderliness in redo log (pingcap#11117) (pingcap#11180) close pingcap#11096 Signed-off-by: qupeng <qupeng@pingcap.com>
What problem does this PR solve?
Issue Number: close #11219
What is changed and how it works?
After #11030, we introduce a mechanism to get the current timestamp
thresholdTS
from pd when changefeed starts, and split all update kv entries whichcommitTS
is smaller than thethresholdTS
.This mechanism has the following problem:
A
andB
, andB
start beforeA
, that isthresholdTSB
<thresholdTSA
;t
is first onA
;t
has an update event whichcommitTS
is smaller thanthresholdTSA
and larger thanthresholdTSB
. So the update event is split to a delete event and an insert event on nodeA
;B
, the update event are received by nodeB
again;B
because itscommitTS
is larger than thethresholdTSB
, and nodeB
just send an update sql to downstream which cause data inconsistency;And there is also another thing to notice that after scheduling, node
B
will send some events to downstream which are already send by nodeA
; So nodeB
must send these events in an idempotent way;Previously, this is handled by getting a
replicateTS
in sink module when sink starts and split these events whichcommitTS
is smaller thanreplicateTS
. But this mechanism is also removed in #11030. So we need to handle this case in puller too.In this pr, instead of maintaining a separate
thresholdTS
in sourcemanager, we try to get thereplicateTS
from sink when puller need to check whether to split the update event.And since puller module starts working before sink module, so we give
replicateTS
a default valueMAXUInt64
which means to split all update events. After sink starts working,replicateTS
will be set to the correct value.The last thing to notice, when sink restarts due to some error, after restart, the sink may send some events downstream which are already send before restart. These events also need be send in an idempotent way. But these events are already in sorter, so just restart sink cannot accomplish this goal. So we forbid restarting sink in this pr and just restart the changefeed when meet error.
Check List
Tests
Questions
Will it cause performance regression or break compatibility?
Do you need to update user documentation, design documentation or monitoring documentation?
Release note