[rabbitmq-discuss] Rabbit won't restart - no next heap size found

[Mike Khristo]

Added a 200G swap... it used some of it, but still seg faults at
startup. Is there any way to split up the data on disk so I can get 2
rabbits up with half the data each?

Any other ideas on how to start this? I need urgent access to the 300M
queued messages

Thanks for the help so far...



On Tue, Sep 25, 2012 at 3:07 PM, Matthias Radestock
<matthias at rabbitmq.com> wrote:
> Mike,
>
>
> On 25/09/12 22:10, Mike Khristo wrote:
>>
>> Upgraded to erlang 5.9.2.
>>
>> eheap_alloc: Cannot allocate 11037567056 bytes of memory (of type
>> "heap").
>
>
> That's better. Sort of.
>
>
>> How can I get rabbit to start w/o using so much memory?
>
>
> You probably can't.
>
>
>> It was only using 30G out of the 68G avail while it was running
>
>
> That is more than the 0.4*total_memory default high watermark. Presumably
> you have altered that threshold in the config. There's a reason it's set to
> 0.4 by default - rabbit can take up all the rest temporarily in some
> circumstances. So you were living on the edge somewhat.
>
>
>> This is a production box and I need to get access to the 300M
>> messages asap.
>
>
> I suggest you configure some temporary swap space. That should hopefully
> allow rabbit to start up and recover all the messages. And, with any luck,
> once it's done all that memory usage might drop off a bit.
>
> Would be good to keep an eye on memory use during that startup/recovery
> phase and send us that plus the log files. And if you have some disk space
> to spare where you can stash away a copy of the database dir before you
> start rabbit that would be great.
>
> With all that info we might be able to identify whether we are missing a
> trick somewhere in conserving memory during startup, and address that in a
> future release.
>
>
> Regards,
>
> Matthias.



-- 
-mk