[vwnc] Replicator failing in VW7.6

I'm using the StoreForGlorpVWUI parcel that comes with VW7.6.  I get the
following error:

Replicating YourPackage-Tests((2.122,dirk),dirk) starting at 6:55:24 am
October 28, 2009 6:59:27.792 - Too much time elapsed. Assuming the
system is hung and quitting.

Now clearly only 4 minutes have elapsed.  This error appears
intermittently.  I checked the code in
StoreAutoReplicator>>startQuitIfReplicationIsStuckLoop and it looks like
its supposed to wait for an hour before giving this error.  I don't see
the bug.  Has anyone else dealt with this?  For now I've just disabled
the "stuck detection".

David

_______________________________________________
vwnc mailing list
[hidden email]
http://lists.cs.uiuc.edu/mailman/listinfo/vwnc

That is odd, and not something I've seen before. I'd guess that somehow that loop got out of sync, but I can't see quite how. Disabling it shouldn't be a problem, unless your replicator frequently does get stuck.

At 10:24 AM 2009-10-28, C. David Shaffer wrote:

I'm using the StoreForGlorpVWUI parcel that comes with VW7.6.  I get the
following error:

Replicating YourPackage-Tests((2.122,dirk),dirk) starting at 6:55:24 am
October 28, 2009 6:59:27.792 - Too much time elapsed. Assuming the
system is hung and quitting.

Now clearly only 4 minutes have elapsed.  This error appears
intermittently.  I checked the code in
StoreAutoReplicator>>startQuitIfReplicationIsStuckLoop and it looks like
its supposed to wait for an hour before giving this error.  I don't see
the bug.  Has anyone else dealt with this?  For now I've just disabled
the "stuck detection".

David

_______________________________________________
vwnc mailing list
[hidden email]
http://lists.cs.uiuc.edu/mailman/listinfo/vwnc

_______________________________________________
vwnc mailing list
[hidden email]
http://lists.cs.uiuc.edu/mailman/listinfo/vwnc

Alan Knight wrote:
> That is odd, and not something I've seen before. I'd guess that
> somehow that loop got out of sync, but I can't see quite how.
> Disabling it shouldn't be a problem, unless your replicator frequently
> does get stuck.
>
Funny you should mention it.  It seemed to slog its way through the
replications I previously had problems with, but as of yesterday it is
getting stuck replicating a bundle.  I can provide more details but
wanted to explore a little myself first because I really count on this
tool.  It looks like it is replicating the individual packages in the
bundle without problems but when it comes time to actually replicate the
bundle it freezes.  Neither postgres nor VW are consuming much CPU
during this waiting period.  Must be network traffic?  It only seems to
happen on bundle boundaries though.  My office machine completes the
replication of said bundle in 3200 seconds (and a top-level bundle which
contains this bundle is still running right now...over 45 minutes)
whereas my much slower laptop (on a narrow pipe) never seems to
finish...or I've never waited long enough.  Any clue why a bundle would
cause such large network load?  Is it a transaction boundary that
contains the package changes too?

David

_______________________________________________
vwnc mailing list
[hidden email]
http://lists.cs.uiuc.edu/mailman/listinfo/vwnc

At 10:32 AM 2009-10-29, C. David Shaffer wrote:

Alan Knight wrote:
> That is odd, and not something I've seen before. I'd guess that
> somehow that loop got out of sync, but I can't see quite how.
> Disabling it shouldn't be a problem, unless your replicator frequently
> does get stuck.
>
Funny you should mention it.  It seemed to slog its way through the
replications I previously had problems with, but as of yesterday it is
getting stuck replicating a bundle.  I can provide more details but
wanted to explore a little myself first because I really count on this
tool.  It looks like it is replicating the individual packages in the
bundle without problems but when it comes time to actually replicate the
bundle it freezes.  Neither postgres nor VW are consuming much CPU
during this waiting period.  Must be network traffic?  It only seems to
happen on bundle boundaries though.  My office machine completes the
replication of said bundle in 3200 seconds (and a top-level bundle which
contains this bundle is still running right now...over 45 minutes)
whereas my much slower laptop (on a narrow pipe) never seems to
finish...or I've never waited long enough.  Any clue why a bundle would
cause such large network load?  Is it a transaction boundary that
contains the package changes too?

There was something that caused bundle replication to be much slower than it needed to be in the 7.6 version, though I think it would have involved it doing far more database reads than it needed to. But it might also have resulted in it using up far more memory than it needed to. Try loading up the latest and see if makes a difference. You might also try increasing the memory sizes, in case it's thrashing there.



_______________________________________________
vwnc mailing list
[hidden email]
http://lists.cs.uiuc.edu/mailman/listinfo/vwnc