Vector events in disk buffer not preserved after pod restart #375
-
When restarting pod with disk buffer persistence enabled vector buffer doesn't try to send already got events from kafka but starts from 0 with current consumer, so events are lost. How this issue could be resolved or I missed something? Use 0.31.0 chart version (the same issue was on 0.26.0 version too, so bump didn't help) Persistence is enabled:
Typical sink config.
P.S. Using Aggregator role with 1 replica |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 5 replies
-
Nothing immediately jumps out. Have you verified that when the pod is restarted, the volume is reattached with all of the disk buffer files contained? These would be in |
Beta Was this translation helpful? Give feedback.
Ah, you are right, I forgot the data directory is mounted in a different directory by the chart. This is to workaround the fact that
/var/lib
is mounted as a read-only filesystem as mentioned by #226 (comment)