[erlang-questions] mnesia_loader overrun by mnesia_table_events

Per Melin per.melin@REDACTED
Mon May 12 14:12:43 CEST 2008


Oops. I only sent my replies to Ulf. See below.

> 2008/5/12 Ulf Wiger <ulf@REDACTED>:
>> One solution that comes to mind is to use transactions rather than
dirty writes.
>
> I tried that, and then the copying works perfect, but Mnesia also
> can't keep up with the writes on the working node in the first place.
> Even if I collect the updates first and write 5000 per transaction and
> use sticky locking.
>
> Hold on. I just tried an explicit full table lock in the transaction.
> That seems to do the trick.
>
> I'll do more testing. If you don't hear anything else this was the
> solution. Thanks!

No, I'm back to square one. Now that Mnesia on the working node (A)
can keep up with the writes then the node coming back up (B) can't
keep up with the copying. It looks like a single transaction of 5000
writes is still sent as 5000 individual table events?



More information about the erlang-questions mailing list