[
https://issues.apache.org/jira/browse/SOLR-7381?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14491655#comment-14491655
]
Shalin Shekhar Mangar edited comment on SOLR-7381 at 4/12/15 7:06 PM:
----------------------------------------------------------------------
Copying over some comments from SOLR-6673:
[[email protected]]'s comment:
https://issues.apache.org/jira/browse/SOLR-6673?focusedCommentId=14381847&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14381847
{quote}
There is some other redundant info now though: C17 P62757
That's the core number and the port number. Both where important in the past
(along with the thread number) to figuring out chaos monkey test fails. But
they don't seem important for production? The port number should be different
for diff log files in production, and the collection name we already have.
{quote}
The log snippet that I pasted in SOLR-6673 was actually a test log which is why
it had C17 and such things. Those are not logged in production logs. There is
still merit in logging the core short id in tests because it can be used to
differentiate between reloaded cores. Now that I am logging the node name, we
don't need the port in the logs so I'll remove it (it is currently logged only
in cloud mode).
[[email protected]]'s comment at
https://issues.apache.org/jira/browse/SOLR-6673?focusedCommentId=14381918&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14381918
{quote}
I don't really like using caps for the label distinction. Seems nicer to have
like c:collection n:core1 or something. Looks a lot nicer, doesn't overload the
c char. I don't really know that it screams "small c, must be core" anyway. We
just need a consistent clean format that is easy script parsable. Or
c:collection s:core1 for the more proper SolrCore?
{quote}
If we use "s:core1" then what should we use for the shard name?
was (Author: shalinmangar):
Copying over some comments from SOLR-6673:
[[email protected]]'s comment:
https://issues.apache.org/jira/browse/SOLR-6673?focusedCommentId=14381847&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14381847
{quote}
There is some other redundant info now though: C17 P62757
That's the core number and the port number. Both where important in the past
(along with the thread number) to figuring out chaos monkey test fails. But
they don't seem important for production? The port number should be different
for diff log files in production, and the collection name we already have.
{quote}
The log snippet that I pasted in SOLR-6673 was actually a test log which is why
it had C17 and such things. Those are not logged in production logs. There is
still merit in logging the core short id in tests because it can be used to
differentiate between reloaded cores. Now that I am logging the node name (at
least for cloud) we don't need the port in the logs so I'll remove it (but only
for cloud).
[[email protected]]'s comment at
https://issues.apache.org/jira/browse/SOLR-6673?focusedCommentId=14381918&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14381918
{quote}
I don't really like using caps for the label distinction. Seems nicer to have
like c:collection n:core1 or something. Looks a lot nicer, doesn't overload the
c char. I don't really know that it screams "small c, must be core" anyway. We
just need a consistent clean format that is easy script parsable. Or
c:collection s:core1 for the more proper SolrCore?
{quote}
If we use "s:core1" then what should we use for the shard name?
> Improve logging by adding MDC context in more places
> ----------------------------------------------------
>
> Key: SOLR-7381
> URL: https://issues.apache.org/jira/browse/SOLR-7381
> Project: Solr
> Issue Type: Improvement
> Components: SolrCloud
> Reporter: Shalin Shekhar Mangar
> Assignee: Shalin Shekhar Mangar
> Priority: Critical
> Fix For: 5.2
>
> Attachments: SOLR-7381.patch
>
>
> SOLR-6673 added MDC based logging in a few places but we have a lot of ground
> to cover. Threads created via thread pool executors do not inherit MDC values
> and those are some of the most interesting places to log MDC context. This is
> critical to help debug SolrCloud failures.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
