[
https://issues.apache.org/jira/browse/IGNITE-14497?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Maria Makedonskaya updated IGNITE-14497:
----------------------------------------
Description:
In [1] has been described scenario of the race that makes possible a situation
when information from metastore has been deleted and part of the index is still
present in the storage.
Actually, fix in [1] that has been done in slightly different way, however, it
strongly decreases possibility of the race. Removal of the background task has
been moved to beginning of the next checkpoint. It was done this way, because
the problem was in case when node unexpectedly failed in the middle of next
checkpoint and after restart task failed on check of deleted index metaPage.
The check was done in [2].
In current issue, we only move background task deletion at the end of the next
checkpoint.
[1][IGNITE-13382|https://issues.apache.org/jira/browse/IGNITE-13382]
[2][IGNITE-14447|https://issues.apache.org/jira/browse/IGNITE-14447]
was:
In [1] has been described scenario of the race that makes possible a situation
when information from metastore has been deleted and part of the index is still
present in the storage.
Actually, fix in [1] that has been done in slightly different way, however, it
strongly decreases possibility of the race. Removal of the background task has
been moved to beginning of the next checkpoint. It was done this way, because
the problem was in case when node unexpectedly failed in the middle of next
checkpoint and after restart task failed on check of deleted index metaPage.
The check was done in [2].
In current issue, we only move background task deletion at the end of the next
checkpoint.
[1]IGNITE-13382
[2]IGNITE-14447
> Move background task deletion at the end of next checkpoint
> -----------------------------------------------------------
>
> Key: IGNITE-14497
> URL: https://issues.apache.org/jira/browse/IGNITE-14497
> Project: Ignite
> Issue Type: Bug
> Reporter: Maria Makedonskaya
> Assignee: Maria Makedonskaya
> Priority: Major
>
> In [1] has been described scenario of the race that makes possible a
> situation when information from metastore has been deleted and part of the
> index is still present in the storage.
> Actually, fix in [1] that has been done in slightly different way, however,
> it strongly decreases possibility of the race. Removal of the background task
> has been moved to beginning of the next checkpoint. It was done this way,
> because the problem was in case when node unexpectedly failed in the middle
> of next checkpoint and after restart task failed on check of deleted index
> metaPage. The check was done in [2].
> In current issue, we only move background task deletion at the end of the
> next checkpoint.
> [1][IGNITE-13382|https://issues.apache.org/jira/browse/IGNITE-13382]
> [2][IGNITE-14447|https://issues.apache.org/jira/browse/IGNITE-14447]
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
