[
https://issues.apache.org/jira/browse/IGNITE-21236?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17813955#comment-17813955
]
Vladimir Steshin commented on IGNITE-21236:
-------------------------------------------
It fails also for the caches filled with the cache API. Not only with the
streamer. The problem seems to rise from wierd initialization
_GridCacheVersionManager#last_:
{code:java}
public void onLocalJoin(long topVer) {
long startTime = cctx.kernalContext().discovery().gridStartTime();
if (startTime != 0)
offset = (int)((startTime - TOP_VER_BASE_TIME) / 1000);
last = new GridCacheVersion(0, order.get(), 0, dataCenterId);
...
}
{code}
It ignores _topVer_, _nodeOrder_ and _offset_. While _GridCacheVersion_ says
'_topVer Topology version plus number of seconds from the start time of the
first grid node._'
> Handle DataStreamer entry version in dump
> -----------------------------------------
>
> Key: IGNITE-21236
> URL: https://issues.apache.org/jira/browse/IGNITE-21236
> Project: Ignite
> Issue Type: Improvement
> Reporter: Maksim Timonin
> Assignee: Vladimir Steshin
> Priority: Major
> Labels: IEP-109, ise
> Attachments: IgniteCacheDumpSelfTest2.java
>
> Time Spent: 0.5h
> Remaining Estimate: 0h
>
> DataStreamer writes entries with fixed GridCacheVersion, that is greater than
> any existing version.
> During dump creation it filters all such entries by version, and no entries
> created by data streamer lands into the dump.
> Possible solution is to apply all entries with this DataStreamer version to
> dump. Other hand, it might make the dump inconsistent when it's created
> parallel to DataStreamer activity (we have the warn for this case in
> Snapshots already)
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
