lucene-solr-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Fran├žois Schiettecatte <>
Subject Re: drastic performance decrease with 20 cores
Date Mon, 26 Sep 2011 16:43:41 GMT
You have not said how big your index is but I suspect that allocating 13GB for your 20 cores
is starving the OS of memory for caching file data. Have you tried 6GB with 20 cores? I suspect
you will see the same performance as 6GB & 10 cores.

Generally it is better to allocate just enough memory to SOLR to run optimally rather than
as much as possible. 'Just enough' depends as well. You will need to try out different allocations
and see where the sweet spot is.



On Sep 26, 2011, at 9:53 AM, Bictor Man wrote:

> Hi everyone,
> Sorry if this issue has been discussed before, but I'm new to the list.
> I have a solr (3.4) instance running with 20 cores (around 4 million docs
> each).
> The instance has allocated 13GB in a 16GB RAM server. If I run several sets
> of queries sequentially in each of the cores, the I/O access goes very high,
> so does the system load, while the CPU percentage remains always low.
> It takes almost 1 hour to complete the set of queries.
> If I stop solr and restart it with 6GB allocated and 10 cores, after a bit
> the I/O access goes down and the CPU goes up, taking only around 5 minutes
> to complete all sets of queries.
> Meaning that for me is MUCH more performant having 2 solr instances running
> with half the data and half the memory than a single instance will all the
> data and memory.
> It would be even way faster to have 1 instance with half the cores/memory,
> run the queues, shut it down, start a new instance and repeat the process
> than having a big instance running everything.
> Furthermore, if I take the 20cores/13GB instance, unload 10 of the cores,
> trigger the garbage collector and run the sets of queries again, the
> behavior still remains slow taking like 30 minutes.
> am I missing something here? does solr change its caching policy depending
> on the number of cores at startup or something similar?
> Any hints will be very appreciated.
> Thanks,
> Victor

View raw message