On Tue, May 17, 2011 at 7:51 PM, Cristi Starasciuc <[email protected]>wrote:

>  On 05/06/2011 07:20 PM, Mircea Carasel wrote:
>
> Hi,
>
>  I am posting a functional patch for XX-9554. GUI is not polished, but
> looks stable to me and shows accurate results
> Feel free to comment the implementation. It took longer than expected to
> bring it in this form, but I was impeded by many other tasks.
>
>  George, Cristi, Douglas - your feedback will be appreciated (if anyone of
> you has time to have a look)
>
>  Here is the patch description:
>
>  -Created cache to hold successful/failed replications
> (FailedReplicationState.java)
> AuditLogContext is responsible with maintaining this cahce
> -created worker in AuditLogContext to push replication finished event
>
>  -saved replication in DB, created two tables linked with location table
> 1-m
> to hold successful/failed replicaitons
> location_success_replications,location_failed_replications
>
>  Mongo replication is saved in one single entry in DB named: Database
> generation
> Other services configuration files, and others names are saved
> Overall there are around 50 records for each location
>
>  UI report, click on server show replication result:
> send profiles shows replication progress
> otherwise we show last replication result from db
> the cache is filled with last replication result during firstrun - in this
> way any ocasional replication fits correcly in chace
> this cache show always the current result of replications
> -save in locaion table data of last replication and status:
> CONFIGURED -replication successful
> CONFIGURATION_ERROR - something failed
>
>  Thank you
> Mircea
>
>
> _______________________________________________
> sipx-dev mailing [email protected]
> List Archive: http://list.sipfoundry.org/archive/sipx-dev/
>
>  Mircea, all
>
> attached is the rebased patch.
>
> I looked at the patch it looks good, seems like you're on the right track
> with this. Here's my input:
>
> I can't see "In progress" state. If I hit Send profiles, the status does
> not change. I think it should be "In progress".
> I would include "Last attempt" field in Replication details page (See state
> link) and not in locations table. Can't see much use for it there.
>
Yes, correct - we should add this new state

> And also, I would put instead of a timestamp a start and end time. (liek in
> Job status, but global)
>
That's a bit hard to accomplish since some files are replicated
asynchronously. One way would be to compare the results saved after
first-run-task, but send profiles does not replicate the same set of files
as first-run-task. sipxconfig-report for instance is replicated only at
first-run task or when a location is updated. Maybe it would be better for
send profiles to replicate everything, as first-run-task does

>
> If I stop Mongo and go to a user, make a change and hit Apply, I get a
> "Replication error" in users' page. Then, if I go to server's page, i will
> see the server is in Configuration error state. Which is not quite
> correct, IMO, since a mongo error would not render the PG and Mongo
> databases out of sync. So there's nothing wrong with the configuration of
> the server. Sure, it will not work as expected, since Mongo being down it
> means chaos to the system.
> If I start Mongo, and go to the user, make the changes and then save, the
> server remains in Configuration error status.
>
> the only way to bring it to Configured state is to send profiles, which in
> the described case is not necessary.
>
> I'd propose that single Mongo replication errors would not produce
> "Configuration error" status; I'm open to discussions, of course.
>
we can put a dedicated state to reflect that the only problem is Mongo is
down

>
> With files is a bit different, a failed file generation can render the file
> not usable, it may have 0 length, or be a malformed xml - that definitely is
> a misconfiguration that may have an impact on certain sipXecs services.
>
> Question: I'm wondering if we could "remember" the failed jobs and have the
> possibility to "retry" them?
>
failed jobs are saved in DB, There are two tables there:
location_failed_replications and location_success_replications where failed
and succesful replicated entity names are saved

Mircea

>
>
> Cheers,
> Cristi
>
> _______________________________________________
> sipx-dev mailing list
> [email protected]
> List Archive: http://list.sipfoundry.org/archive/sipx-dev/
>
_______________________________________________
sipx-dev mailing list
[email protected]
List Archive: http://list.sipfoundry.org/archive/sipx-dev/

Reply via email to