[jira] [Commented] (CASSANDRA-13313) Compaction leftovers not removed on upgrade 2.1/2.2 -> 3.0

2017-12-19 Thread Jeff Jirsa (JIRA)

[ 
https://issues.apache.org/jira/browse/CASSANDRA-13313?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16297215#comment-16297215
 ] 

Jeff Jirsa commented on CASSANDRA-13313:


Reopening, patch coming soon.


> Compaction leftovers not removed on upgrade 2.1/2.2 -> 3.0
> --
>
> Key: CASSANDRA-13313
> URL: https://issues.apache.org/jira/browse/CASSANDRA-13313
> Project: Cassandra
>  Issue Type: Bug
>  Components: Compaction
>Reporter: Jeff Jirsa
>Assignee: Jeff Jirsa
>Priority: Minor
>
> Before 3.0 we used sstable ancestors to figure out if an sstable was left 
> over after a compaction. In 3.0 the ancestors are ignored and instead we use 
> LogTransaction files to figure it out. 3.0 should still clean up 2.1/2.2 
> compaction leftovers using the on-disk sstable ancestors when available.



--
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



[jira] [Commented] (CASSANDRA-13313) Compaction leftovers not removed on upgrade 2.1/2.2 -> 3.0

2017-03-23 Thread Stefania (JIRA)

[ 
https://issues.apache.org/jira/browse/CASSANDRA-13313?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15939531#comment-15939531
 ] 

Stefania commented on CASSANDRA-13313:
--

> My expectation was the only consistency issue is counters 

Counters created before 2.1, after 2.1 it is no longer a concern.

> there are other resource issues (disk space, memory on reads if there are a 
> lot of tombstones, etc).

Correct, during the 7066 discussions, I remember this was mentioned, that at 
worst it would be a resource issue and only affecting users who would attempt 
to upgrade without a clean shutdown.

> I'm not sure how vital it is - I marked it minor and it's low on my queue, 
> but if consensus is it's truly a won't-fix I'm not sure I'll fight that.

CASSANDRA-12212 was a won't fixed mostly because the benefit was very low. 
However, if there is a patch, I am not at all opposed to it.

> Compaction leftovers not removed on upgrade 2.1/2.2 -> 3.0
> --
>
> Key: CASSANDRA-13313
> URL: https://issues.apache.org/jira/browse/CASSANDRA-13313
> Project: Cassandra
>  Issue Type: Bug
>  Components: Compaction
>Reporter: Jeff Jirsa
>Assignee: Jeff Jirsa
>Priority: Minor
>
> Before 3.0 we used sstable ancestors to figure out if an sstable was left 
> over after a compaction. In 3.0 the ancestors are ignored and instead we use 
> LogTransaction files to figure it out. 3.0 should still clean up 2.1/2.2 
> compaction leftovers using the on-disk sstable ancestors when available.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)


[jira] [Commented] (CASSANDRA-13313) Compaction leftovers not removed on upgrade 2.1/2.2 -> 3.0

2017-03-23 Thread Jeff Jirsa (JIRA)

[ 
https://issues.apache.org/jira/browse/CASSANDRA-13313?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15939519#comment-15939519
 ] 

Jeff Jirsa commented on CASSANDRA-13313:


Oh I hadn't seen 12212

I have a patch where it's implemented with {{compactions_in_progress}} and 
re-added (but deprecated) ancestors to metadata , but hadn't finished the dtest.

My expectation was the only consistency issue is counters but there are other 
resource issues (disk space, memory on reads if there are a lot of tombstones, 
etc).

I'm not sure how vital it is - I marked it minor and it's low on my queue, but 
if consensus is it's truly a won't-fix I'm not sure I'll fight that.



> Compaction leftovers not removed on upgrade 2.1/2.2 -> 3.0
> --
>
> Key: CASSANDRA-13313
> URL: https://issues.apache.org/jira/browse/CASSANDRA-13313
> Project: Cassandra
>  Issue Type: Bug
>  Components: Compaction
>Reporter: Jeff Jirsa
>Assignee: Jeff Jirsa
>Priority: Minor
>
> Before 3.0 we used sstable ancestors to figure out if an sstable was left 
> over after a compaction. In 3.0 the ancestors are ignored and instead we use 
> LogTransaction files to figure it out. 3.0 should still clean up 2.1/2.2 
> compaction leftovers using the on-disk sstable ancestors when available.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)


[jira] [Commented] (CASSANDRA-13313) Compaction leftovers not removed on upgrade 2.1/2.2 -> 3.0

2017-03-23 Thread Stefania (JIRA)

[ 
https://issues.apache.org/jira/browse/CASSANDRA-13313?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15939503#comment-15939503
 ] 

Stefania commented on CASSANDRA-13313:
--

There was a similar discussion in CASSANDRA-12212 - albeit focused on using 
{{compactions_in_progress}} rather than ancestors. Other than counters created 
before 2.1, are you aware of any other possible consistency issues?

> Compaction leftovers not removed on upgrade 2.1/2.2 -> 3.0
> --
>
> Key: CASSANDRA-13313
> URL: https://issues.apache.org/jira/browse/CASSANDRA-13313
> Project: Cassandra
>  Issue Type: Bug
>  Components: Compaction
>Reporter: Jeff Jirsa
>Assignee: Jeff Jirsa
>Priority: Minor
>
> Before 3.0 we used sstable ancestors to figure out if an sstable was left 
> over after a compaction. In 3.0 the ancestors are ignored and instead we use 
> LogTransaction files to figure it out. 3.0 should still clean up 2.1/2.2 
> compaction leftovers using the on-disk sstable ancestors when available.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)