Skip to content

Transactions get stuck sometimes #28768

@kirillvasilenko

Description

@kirillvasilenko

You may see it on the jepsen latency chart like this

Image

It does not causes anomalies, but it makes the latency very bad. So we need to fix that.

Probable causes:

  1. There may be a problem in the distributed commit logic. In some cases, the events come in such an order that the transaction does not progress and stuck. TEvWakeUp comes after the timeout, and wakes the tx up, so it progresses and finishes.
  2. Some heavy system transactions may coma and just take too much time to execute, so all the others wait.

Important note: this problem can be reproduced even on a table of 3 partitions (that makes reading the logs and debugging much easier)

Metadata

Metadata

Labels

No labels
No labels

Type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions