[
https://issues.apache.org/jira/browse/CASSANDRA-20581?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17951916#comment-17951916
]
Jaydeepkumar Chovatia commented on CASSANDRA-20581:
---------------------------------------------------
I will create a separate PR for the following suggestion from [~clohfink]
"maybe a small addition - can you add a "unrepairedAge" in TableMetrics or the
like which gives oldest mutation (from sstable metadata) from the unrepaired
set? Just sstable age isnt always sufficient cause hints/repair streaming can
sneak older stuff in."
{code:java}
unrepairedAge = createTableGauge("UnrepairedAgeInSeconds", () ->
{
long oldest = Long.MAX_VALUE;
for (SSTableReader sstable : cfs.getSSTables(SSTableSet.CANONICAL))
{
if (!sstable.isRepaired())
{
oldest = Math.min(oldest, sstable.getMinTimestamp());
}
}
return Math.max(0, FBUtilities.nowInSeconds() -
TimeUnit.MICROSECONDS.toSeconds(oldest));
} {code}
> Improved observability in AutoRepair to report both expected vs. actual
> repair bytes and expected vs. actual keyspaces
> ----------------------------------------------------------------------------------------------------------------------
>
> Key: CASSANDRA-20581
> URL: https://issues.apache.org/jira/browse/CASSANDRA-20581
> Project: Apache Cassandra
> Issue Type: Task
> Components: Consistency/Repair
> Reporter: Jaydeepkumar Chovatia
> Assignee: Jaydeepkumar Chovatia
> Priority: Normal
> Time Spent: 10m
> Remaining Estimate: 0h
>
> The current AutoRepair framework has enough visibility to know which nodes
> are going through repair, the duration they ran repair, the token ranges
> repaired, the failed ranges, and the successful/failed keyspaces/tables.
> However, on a node when repair is running, we miss a few fine-grained signals
> {*}while repair is in progress{*}, such as the following:
> # Number of tables/keyspaces/token ranges repaired vs. pending
> # Number of Merkle trees/token-ranges out of sync, which indirectly tells us
> inconsistencies among the nodes
> ## Convert this metric into % of data in sync vs. not
>
> Fine-grained node-level observability also came up while
> [reviewing|https://github.com/jaydeepkumar1984/cassandra/pull/54#issuecomment-2813945023]
> the POC for Repair on bootstrap, as it is a must for repairing as part of
> the bootstrap.
> This ticket is to improve the AutoRepair observability, and make it similar
> to _nodetool compactionstats_
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org
