Login | Register For Free | Help
Search for: (Advanced)

Mailing List Archive: Lucene: Java-Dev

[jira] [Commented] (SOLR-3591) Startup error not reflected in Solr web view

 

 

Lucene java-dev RSS feed   Index | Next | Previous | View Threaded


jira at apache

Jul 3, 2012, 11:32 PM

Post #1 of 5 (85 views)
Permalink
[jira] [Commented] (SOLR-3591) Startup error not reflected in Solr web view

[ https://issues.apache.org/jira/browse/SOLR-3591?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13406300#comment-13406300 ]

Stefan Matheis (steffkes) commented on SOLR-3591:
-------------------------------------------------

Just for reproducing the UI-Steps, i changed {{elevate.xml}} into {{elevate.xm}} to produce an exception:

* We're trying to get Information about available cores from {{/solr/admin/cores?wt=json}}:
{code}{
responseHeader: {
status: 0,
QTime: 1
},
defaultCoreName: "collection1",
status: { }
}{code}

* Afterwards we try to fetch System Information (from the first available core), because there are no cores listed .. we end up with {{/solr/null/admin/system?wt=json}} - which just gives an {{404 Not Found}}

So how to decide if the user has no admin-handler defined or we have an startup error? my first idea was to display the latest logentry when we display this error message, so the user may decide himself what happend - but this will not work because {{/solr/admin/logging}} also responds with a {{404 Not Found}}

Right now i don't really have an idea how to solve that?

> Startup error not reflected in Solr web view
> --------------------------------------------
>
> Key: SOLR-3591
> URL: https://issues.apache.org/jira/browse/SOLR-3591
> Project: Solr
> Issue Type: Bug
> Components: web gui
> Affects Versions: 4.0-ALPHA
> Reporter: Erik Hatcher
> Assignee: Stefan Matheis (steffkes)
> Priority: Blocker
> Fix For: 4.0
>
> Attachments: screenshot-1.jpg
>
>
> When Solr has a fatal startup error, it used to be reflected in general responses from Solr. With the new UI, it's relegated to only the logs.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira



---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe [at] lucene
For additional commands, e-mail: dev-help [at] lucene


jira at apache

Jul 4, 2012, 1:10 PM

Post #2 of 5 (79 views)
Permalink
[jira] [Commented] (SOLR-3591) Startup error not reflected in Solr web view [In reply to]

[ https://issues.apache.org/jira/browse/SOLR-3591?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13406706#comment-13406706 ]

Stefan Matheis (steffkes) commented on SOLR-3591:
-------------------------------------------------

hum, quick thought was, that we maybe see a difference in the output if the AdminCoreHandler .. but that's not the fact .. it looks exactly the same. having no cores defined gives the same result as an startup error.

Is there really no way to see/detect if solr was started like expected? Until that Issue i was pretty sure, that we have most of the cases covered ;o

Perhaps someone can give a few insights, why all the resources return 404 Not Found even w/o hitting Solr itself? At least you'll see no requests in the default logs .. which are normally populated for every request.

> Startup error not reflected in Solr web view
> --------------------------------------------
>
> Key: SOLR-3591
> URL: https://issues.apache.org/jira/browse/SOLR-3591
> Project: Solr
> Issue Type: Bug
> Components: web gui
> Affects Versions: 4.0-ALPHA
> Reporter: Erik Hatcher
> Assignee: Stefan Matheis (steffkes)
> Priority: Blocker
> Fix For: 4.0
>
> Attachments: screenshot-1.jpg
>
>
> When Solr has a fatal startup error, it used to be reflected in general responses from Solr. With the new UI, it's relegated to only the logs.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira



---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe [at] lucene
For additional commands, e-mail: dev-help [at] lucene


jira at apache

Jul 10, 2012, 3:16 PM

Post #3 of 5 (76 views)
Permalink
[jira] [Commented] (SOLR-3591) Startup error not reflected in Solr web view [In reply to]

[ https://issues.apache.org/jira/browse/SOLR-3591?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13410991#comment-13410991 ]

Hoss Man commented on SOLR-3591:
--------------------------------

It sounds like there are two related problems that compound int oa really bad experience...

----

*1) the web ui isn't cleaning dealing with the situation where there are "no cores" returned by the CoreAdminHandler.*

this is a legitimate situation, that doesn't neccessarily indicate an error.

if there are no cores, then the ui shouldn't blindly try to load "/solr/null/admin/system?wt=json" and then complain that the admin handler can't be found -- the logic should be something like:

* can we talk to CoreAdminHandler at all? if not give a specific error
* if CoreAdminHandler says there are no cores, give a message to that effect
** offer the info/commands that are stil available (ie: "Core Admin" functionality)
** perhaps suggesting that if they expected to cores to already exist, they should check logs 9allthough this may not be needed depending on how far we get with "#2" below)
* if cores are available, then try to use /corename/admin to get the other info to populate the UI, and if we can't find it *then* mention that they need to add config
** i would also suggest using the "defaultcore" if non-null instead of just whatever core happens to be listed first (but that's a good fallback if there is no default core)

----

*2) no external reporting of errors in initializing cores*

once upon a time, Solr had an "abortOnConfigurationError" option per core, that never really worked well, and would try to partially initialize a core even if some things failed. In conjunction with that (broken) setting, was a static list of Exceptions that had been thrown during SolrCore construction, which the SolrRequestDispatcher would try to use to generate an error messages if there was a problem.

All of that code has been ripped out of trunk for a long while, largely because it didn't work, and it _REALLY_ didn't work well with multicore, but as erik points out the one thing that it _did_ help with was in making it obvious when there were config problems on startup.

I think we should at least partially revive the idea of keeping track of the list of "severe" errors, but there are a lot of things we can do differnetly now:

* instead of being static, it can be managed by the CoreContainer
* it can specificly be exceptions caught by CoreContainer while initializing SolrCores.
* we can maintain it as a map of (String)coreName -> Pair<(Date)timstamp,(Exception)error> so it's clear what exception came from initializing which solr core
** by using a name mapping, we can delete entires if/when a SolrCore is re-loaded to fix the error.
* we can return this map in CoreAdminHandler so the UI can display a big flashing warning about cores that failed to initialize (both on startup, or if some remote command tried to create a core programaticly)

----

i suggest we spin off a new Jira for #1 since that is somewhat independent (we need to change the "no cores" behavior of the UI no matter what else we do) and use sub-tasks of this issue to track improvements to CoreContainer/CoreAdminHandler/UI to display errors related to SolrCore initialization.

sound good?

> Startup error not reflected in Solr web view
> --------------------------------------------
>
> Key: SOLR-3591
> URL: https://issues.apache.org/jira/browse/SOLR-3591
> Project: Solr
> Issue Type: Bug
> Components: web gui
> Affects Versions: 4.0-ALPHA
> Reporter: Erik Hatcher
> Assignee: Stefan Matheis (steffkes)
> Priority: Blocker
> Fix For: 4.0
>
> Attachments: screenshot-1.jpg
>
>
> When Solr has a fatal startup error, it used to be reflected in general responses from Solr. With the new UI, it's relegated to only the logs.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira



---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe [at] lucene
For additional commands, e-mail: dev-help [at] lucene


jira at apache

Jul 10, 2012, 3:38 PM

Post #4 of 5 (70 views)
Permalink
[jira] [Commented] (SOLR-3591) Startup error not reflected in Solr web view [In reply to]

[ https://issues.apache.org/jira/browse/SOLR-3591?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13411011#comment-13411011 ]

Ryan McKinley commented on SOLR-3591:
-------------------------------------

Perhaps we can use part of SOLR-3358 to capture FATAL exceptions and somehow expose them regardless of the core?

Then the UI behavior could essentially show the logging when /solr/admin/xxxx does not exist?



> Startup error not reflected in Solr web view
> --------------------------------------------
>
> Key: SOLR-3591
> URL: https://issues.apache.org/jira/browse/SOLR-3591
> Project: Solr
> Issue Type: Bug
> Components: web gui
> Affects Versions: 4.0-ALPHA
> Reporter: Erik Hatcher
> Assignee: Stefan Matheis (steffkes)
> Priority: Blocker
> Fix For: 4.0
>
> Attachments: screenshot-1.jpg
>
>
> When Solr has a fatal startup error, it used to be reflected in general responses from Solr. With the new UI, it's relegated to only the logs.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira



---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe [at] lucene
For additional commands, e-mail: dev-help [at] lucene


jira at apache

Jul 10, 2012, 4:35 PM

Post #5 of 5 (74 views)
Permalink
[jira] [Commented] (SOLR-3591) Startup error not reflected in Solr web view [In reply to]

[ https://issues.apache.org/jira/browse/SOLR-3591?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13411059#comment-13411059 ]

Hoss Man commented on SOLR-3591:
--------------------------------

bq. Perhaps we can use part of SOLR-3358 to capture FATAL exceptions and somehow expose them regardless of the core?

i don't have any objections to that, what i've seen of SOLR-3358 looks great so far, and it would be nice if it was accesible even w/o having any SolrCores -- but i think we need a more specific targeted solution for reporting core initialization problems...

* we can't uniquely identify which core caused which log messages
* we can't control what log messages might come out of a plugin from a core
* we can't identify which log message was the "straw that broke the camels back" and actually caused the core init to fail.
* we can't definitively know if a log message is "still important" as more log messages come in (from other cores)
* we can't know if a specific log messages related to core initialization is "still a problem" or if that specific core has already been fix and re-created

...but we can, in CoreContainer, catch and record the specific exceptions related to each core name, and track them relative that core name, and let CoreAdminHandler have thta data when it's asked to report status.

so if a plugin in coreA logged 99 "fatal" log messages, but coreA still started fine; while coreB didn't log anything but the constructor threw an exception X we can make CoreAdminHAndler reliably (and confidently) say "here's your status for coreA, and FYI: coreB failed to initialize because of X" w/o making the user wade through 100 other log messages that are unrelated. And even if the user doesn't look at the core status for hours and hours after trying to startup (or after some cron tried to programaticly create coreB), and there have been thousands of other "errors" logged by other cores, CoreAdminHandler can still say "this error X is the reason you don't have a coreB right now".

see what i mean?


> Startup error not reflected in Solr web view
> --------------------------------------------
>
> Key: SOLR-3591
> URL: https://issues.apache.org/jira/browse/SOLR-3591
> Project: Solr
> Issue Type: Bug
> Components: web gui
> Affects Versions: 4.0-ALPHA
> Reporter: Erik Hatcher
> Assignee: Stefan Matheis (steffkes)
> Priority: Blocker
> Fix For: 4.0
>
> Attachments: screenshot-1.jpg
>
>
> When Solr has a fatal startup error, it used to be reflected in general responses from Solr. With the new UI, it's relegated to only the logs.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira



---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe [at] lucene
For additional commands, e-mail: dev-help [at] lucene

Lucene java-dev RSS feed   Index | Next | Previous | View Threaded
 
 


Interested in having your list archived? Contact Gossamer Threads
 
  Web Applications & Managed Hosting Powered by Gossamer Threads Inc.