[
https://issues.apache.org/jira/browse/CASSANDRA-19120?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17791722#comment-17791722
]
Jaydeepkumar Chovatia edited comment on CASSANDRA-19120 at 11/30/23 5:44 PM:
-----------------------------------------------------------------------------
I don't think so; this is completely resolved in the _trunk_ yet. We are still
not adjusting the _latch_ if the _pendingRepairs_ include a node from the
remote region.
Let me explain with the following scenario:
# _pendingRepairs_ at line number 81 [1] has 1 node from the local region and
another node from a remote region
# At line number 100 [2] _blockFor_ will be 2, i.e., the latch will be 2
# When the local node responds, then at line no 122 [3], it includes the
response, i.e., the latch will be decremented as
*{color:#00875a}expected{color}*
# When the remote node responds, then at line no 122 [3], it excludes the
response, i.e., the latch will not be {color:#de350b}decremented{color}, and
this is the *{color:#ff0000}problem{color}*
Could you please double-check the above scenario?
[1][https://github.com/apache/cassandra/blob/trunk/src/java/org/apache/cassandra/service/reads/repair/BlockingPartitionRepair.java#L81]
[2][https://github.com/apache/cassandra/blob/trunk/src/java/org/apache/cassandra/service/reads/repair/BlockingPartitionRepair.java#L100]
[3]
[https://github.com/apache/cassandra/blob/trunk/src/java/org/apache/cassandra/service/reads/repair/BlockingPartitionRepair.java#L122]
was (Author: chovatia.jayd...@gmail.com):
I don't think so; this is completely resolved in the _trunk_ yet. We are still
not adjusting the _latch_ if the _pendingRepairs_ include a node from the
remote region.
Let me explain with the following scenario:
# _pendingRepairs_ at line number 81 [1] has 1 node from the local region and
another node from a remote region
# At line number 100 [2] _blockFor_ will be 2, i.e., the latch will be 2
# When the local node responds, then at line no 122 [3], it includes the
response, i.e., the latch will be decremented as
*{color:#00875a}expected{color}*
# When the remote node responds, then at line no 122 [3], it excludes the
response, i.e., the latch will not be {color:#de350b}decremented{color}, and
this is the *{color:#ff0000}problem{color}*
Could you please double-check the above scenario?
[1]https://github.com/apache/cassandra/blob/trunk/src/java/org/apache/cassandra/service/reads/repair/BlockingPartitionRepair.java#L81
[2][https://github.com/apache/cassandra/blob/trunk/src/java/org/apache/cassandra/service/reads/repair/BlockingPartitionRepair.java#L100]
[3]
[https://github.com/apache/cassandra/blob/trunk/src/java/org/apache/cassandra/service/reads/repair/BlockingPartitionRepair.java#L122]
> local consistencies may get timeout if blocking read repair is sending the
> read repair mutation to other DC
> ------------------------------------------------------------------------------------------------------------
>
> Key: CASSANDRA-19120
> URL: https://issues.apache.org/jira/browse/CASSANDRA-19120
> Project: Cassandra
> Issue Type: Bug
> Reporter: Runtian Liu
> Priority: Normal
> Attachments: image-2023-11-29-15-26-08-056.png, signature.asc
>
>
> For a two DCs cluster setup. When a new node is being added to DC1, for
> blocking read repair triggered by local_quorum in DC1, it will require to
> send read repair mutation to an extra node(1)(2). The selector for read
> repair may select *ANY* node that has not been contacted before(3) instead of
> selecting the DC1 nodes. If a node from DC2 is selected, this will cause 100%
> timeout because of the bug described below:
> When we initialized the latch(4) for blocking read repair, the shouldBlockOn
> function will only return true for local nodes(5), the blockFor value will be
> reduced if a local node doesn't require repair(6). The blockFor is same as
> the number of read repair mutation sent out. But when the coordinator node
> receives the response from the target nodes, the latch only count down for
> nodes in same DC(7). The latch will wait till timeout and the read request
> will timeout.
> This can be reproduced if you have a constant load on a 3 + 3 cluster when
> adding a node. If you have someway to trigger blocking read repair(maybe by
> adding load using stress tool). If you use local_quorum consistency with a
> constant read after write load in the same DC that you are adding node. You
> will see read timeout issue from time to time because of the bug described
> above
>
> I think for read repair when selecting the extra node to do repair, we should
> prefer local nodes than the nodes from other region. Also, we need to fix the
> latch part so even if we send mutation to the nodes in other DC, we don't get
> a timeout.
> (1)[https://github.com/apache/cassandra/blob/cassandra-4.0.11/src/java/org/apache/cassandra/locator/ReplicaPlans.java#L455]
> (2)[https://github.com/apache/cassandra/blob/trunk/src/java/org/apache/cassandra/db/ConsistencyLevel.java#L183]
> (3)[https://github.com/apache/cassandra/blob/cassandra-4.0.11/src/java/org/apache/cassandra/locator/ReplicaPlans.java#L458]
> (4)[https://github.com/apache/cassandra/blob/cassandra-4.0.11/src/java/org/apache/cassandra/service/reads/repair/BlockingPartitionRepair.java#L96]
> (5)[https://github.com/apache/cassandra/blob/cassandra-4.0.11/src/java/org/apache/cassandra/service/reads/repair/BlockingPartitionRepair.java#L71]
> (6)[https://github.com/apache/cassandra/blob/cassandra-4.0.11/src/java/org/apache/cassandra/service/reads/repair/BlockingPartitionRepair.java#L88]
> (7)[https://github.com/apache/cassandra/blob/cassandra-4.0.11/src/java/org/apache/cassandra/service/reads/repair/BlockingPartitionRepair.java#L113]
>
--
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
