Job freeze

[ig...@ig.com.br]

After starting a job I see a lot (a burst) of these errors in 
manifoldcf.log file:

ERROR 2015-12-05 17:27:06,181 (Idle cleanup thread) - Exception tossed: 
Service '_ANON_1' of type '_TRANSFORMATIONCONNECTORPOOL_Documentos' is 
not active
org.apache.manifoldcf.core.interfaces.ManifoldCFException: Service 
'_ANON_1' of type '_TRANSFORMATIONCONNECTORPOOL_Documentos' is not 
active

I have that transformation connect in my job pipeline as the first 
precedent, it's an "Allowed documents" type, and his "connection status" 
states: "Connection working"

At the same time, in the pg_log file of postgresql, there is several of 
this errors

Dec 5 17:25:43 manifold postgres[94028]: [1122-1] ERROR:  could not 
serialize access due to read/write dependencies among transactions
Dec 5 17:25:43 manifold postgres[94028]: [1122-2] DETAIL:  Reason code: 
Canceled on identification as a pivot, during write.
Dec 5 17:25:43 manifold postgres[94028]: [1122-3] HINT:  The transaction 
might succeed if retried.
Dec 5 17:25:43 manifold postgres[94028]: [1122-4] STATEMENT:  UPDATE 
jobqueue SET processid=$1,status=$2 WHERE id=$3

After that the job freeze (in "status and job management" the statistics 
stop being refreshed) and I have to abort it.

I'm running manifoldCF and postgresql in the same machine (FreeBSD) to 
feed another machine with solr (FreeBSD too and nothing showing up in 
solr logs)

Does anybody have a hint to solve those errors?

Re: Job freeze

[Karl Wright <da...@gmail.com>]

Hi,

The first problem sounds like it might be related to your multi-process
configuration.  How is this deployed?  I would guess zookeeper plus
multiple agents processes? It's basically complaining that a connection
pool management service that should exist in Zookeeper does not appear to
be active, although it exists.  The only way to get that, I think, is to
have multiple processes that all have the same process identifier.

The second problem may well be related to that.  In any case, it means that
Postgresql cannot in fact execute something within a transaction so that
transaction is getting aborted.  There are many places in MCF where we
catch database aborts of that kind and retry; in order to see exactly where
this is taking place I'd need more of the stack trace.


Thanks,
Karl




On Sat, Dec 5, 2015 at 3:06 PM, <ig...@ig.com.br> wrote:

> After starting a job I see a lot (a burst) of these errors in
> manifoldcf.log file:
>
> ERROR 2015-12-05 17:27:06,181 (Idle cleanup thread) - Exception tossed:
> Service '_ANON_1' of type '_TRANSFORMATIONCONNECTORPOOL_Documentos' is not
> active
> org.apache.manifoldcf.core.interfaces.ManifoldCFException: Service
> '_ANON_1' of type '_TRANSFORMATIONCONNECTORPOOL_Documentos' is not active
>
> I have that transformation connect in my job pipeline as the first
> precedent, it's an "Allowed documents" type, and his "connection status"
> states: "Connection working"
>
> At the same time, in the pg_log file of postgresql, there is several of
> this errors
>
> Dec 5 17:25:43 manifold postgres[94028]: [1122-1] ERROR:  could not
> serialize access due to read/write dependencies among transactions
> Dec 5 17:25:43 manifold postgres[94028]: [1122-2] DETAIL:  Reason code:
> Canceled on identification as a pivot, during write.
> Dec 5 17:25:43 manifold postgres[94028]: [1122-3] HINT:  The transaction
> might succeed if retried.
> Dec 5 17:25:43 manifold postgres[94028]: [1122-4] STATEMENT:  UPDATE
> jobqueue SET processid=$1,status=$2 WHERE id=$3
>
> After that the job freeze (in "status and job management" the statistics
> stop being refreshed) and I have to abort it.
>
> I'm running manifoldCF and postgresql in the same machine (FreeBSD) to
> feed another machine with solr (FreeBSD too and nothing showing up in solr
> logs)
>
> Does anybody have a hint to solve those errors?
>