You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@pegasus.apache.org by GitBox <gi...@apache.org> on 2022/02/08 03:42:24 UTC
[GitHub] [incubator-pegasus] hycdong opened a new issue #902: Bug: can not get last_flushed_decree from meta cf occasionally when rolling_update from 2.0.x
hycdong opened a new issue #902:
URL: https://github.com/apache/incubator-pegasus/issues/902
## Bug Report
When rolling_update Pegasus server from 2.0.0 to 2.2.0, replica server generate a coredump that could not find the last_flushed_decree occasionally.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: dev-unsubscribe@pegasus.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@pegasus.apache.org
For additional commands, e-mail: dev-help@pegasus.apache.org
[GitHub] [incubator-pegasus] hycdong commented on issue #902: Bug: can not get last_flushed_decree from meta cf occasionally when rolling_update from 2.0.x
Posted by GitBox <gi...@apache.org>.
hycdong commented on issue #902:
URL: https://github.com/apache/incubator-pegasus/issues/902#issuecomment-1034674782
# Basic background
In Pegasus 1.x version, each write request will write request-decree into RocksDB Manifest file, which is hard for us to upgrade RocksDB official new release. To remove this logic, Pegasus firstly release 2.0.0, which will write request-decree both into Manifest file and RocksDB column family called `meta`, and read it from Manifest file. In the following release 2.1.0, 2.2.0, decree will only be read and wrote through meta column family.
# When and how will it happend
When the cluster upgrade Pegasus from 2.0.0 to higher version(for example 2.2.0), the bug will be happened satifying all following conditions:
1. Cluster has a table doesn't trigger flush memtable after upgrading to 2.0.0
2. One partition of this table is not close gracefully when upgrading to 2.2.0 (When close function called, partition will flush memtable, inluding meta column famlity, decree would be found while reopen RocksDB instance).
# How to avoid it
To avoid this bug, we should gurantee partitions would be closed gracefully in upgrade process.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: dev-unsubscribe@pegasus.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@pegasus.apache.org
For additional commands, e-mail: dev-help@pegasus.apache.org