[ 
https://issues.apache.org/jira/browse/CASSANDRA-12173?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16089335#comment-16089335
 ] 

Kurt Greaves commented on CASSANDRA-12173:
------------------------------------------

I think the system_traces issue would have been unrelated to MV's. 

It's expected that if you create a MV your ViewWrite will increase proportional 
to how many writes you do to the base table. I'm not sure if just the Write 
count is meant to reflect this as well. Plus also the writes to system_traces 
would have counted towards your overall writes.

What exactly do you think is the bug here? 

> Materialized View may turn on TRACING
> -------------------------------------
>
>                 Key: CASSANDRA-12173
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-12173
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Materialized Views
>            Reporter: Hiroshi Usami
>
> We observed this in our test cluster(C*3.0.6), but TRAING was OFF apparently.
> After creating Materialized View, the Write count jumped up to 20K from 5K, 
> and the ViewWrite rose up to 10K.
> This is supposed to be done by MV, but some nodes which had 14,000+ SSTables 
> in the system_traces directory went down in a half day, because of running 
> out of file descriptors.
> {code}
> Counting by: find /var/lib/cassandra/data/system_traces/ -name "*-Data.db"|wc 
> -l
>       node01: 0
>       node02: 3
>       node03: 1
>       node04: 0
>       node05: 0
>       node06: 0
>       node07: 2
>       node08: 0
>       node09: 0
>       node10: 0
>       node11: 2
>       node12: 2
>       node13: 1
>       node14: 7
>       node15: 1
>       node16: 5
>       node17: 0
>       node18: 0
>       node19: 0
>       node20: 0
>       node21: 1
>       node22: 0
>       node23: 2
>       node24: 14420
>       node25: 0
>       node26: 2
>       node27: 0
>       node28: 1
>       node29: 1
>       node30: 2
>       node31: 1
>       node32: 0
>       node33: 0
>       node34: 0
>       node35: 14371
>       node36: 0
>       node37: 1
>       node38: 0
>       node39: 0
>       node40: 1
> {code}
> In node24, the sstabledump of the oldest SSTable in system_traces/events 
> directory starts with:
> {code}
> [
>   {
>     "partition" : {
>       "key" : [ "e07851d0-4421-11e6-abd7-59d7f275ba79" ],
>       "position" : 0
>     },
>     "rows" : [
>       {
>         "type" : "row",
>         "position" : 30,
>         "clustering" : [ "e07878e0-4421-11e6-abd7-59d7f275ba79" ],
>         "liveness_info" : { "tstamp" : "2016-07-07T09:04:57.197Z", "ttl" : 
> 86400, "expires_at" : "2016-07-08T09:04:57Z", "expired" : true },
>         "cells" : [
>           { "name" : "activity", "value" : "Parsing CREATE MATERIALIZED VIEW
> ...
> {code}
> So this could be the begining of TRACING ON implicitly. In node35, the oldest 
> one also starts with the "Parsing CREATE MATERIALIZED VIEW".



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org

Reply via email to