lucene-solr-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
Subject Re: SolrCloud and LVM
Date Mon, 09 Jan 2017 15:37:34 GMT
Yeah we normally take the number of GB on a machine for the index size on disk and then double
it for memory...

For example we have 28gb on disk and we see great perf at 64gb ram.

If you can do that you will probably get good results. Remember to not give Java much memory.
We set it at 12gb. We call it starving Java and it reduces the time to garbage collect to
small increments.

Bill Bell
Sent from mobile

> On Jan 9, 2017, at 7:56 AM, Chris Ulicny <> wrote:
> That's good to hear. I didn't think there would be any reason that using
> lvm would impact solr's performance but wanted to see if there was anything
> I've missed.
> As far as other performance goes, we use pcie and sata solid state drives
> since the indexes are mostly too large to cache entirely in memory, and we
> haven't had any performance problems so far. So I'm not expecting that to
> change too much when moving the cloud architecture.
> Thanks again.
>> On Thu, Jan 5, 2017 at 7:55 PM Shawn Heisey <> wrote:
>>> On 1/5/2017 3:12 PM, Chris Ulicny wrote:
>>> Is there any known significant performance impact of running solrcloud
>> with
>>> lvm on linux?
>>> While migrating to solrcloud we don't have the storage capacity for our
>>> expected final size, so we are planning on setting up the solrcloud
>>> instances on a logical volume that we can grow when hardware becomes
>>> available.
>> Nothing specific.  Whatever the general performance impacts for LVM are
>> is what Solr would encounter when it reads and writes data to/from the
>> disk.
>> If your system has enough memory for good performance, then disk reads
>> will be rare, so the performance of the storage volume wouldn't matter
>> much.  If you don't have enough memory, then the disk performance would
>> matter ...although Solr's performance at that point would probably be
>> bad enough that you'd be looking for ways to improve it.
>> Here's some information:
>> Exactly how much memory is enough depends on enough factors that there's
>> no good general advice.  The only thing we can say in general is to
>> recommend the ideal setup -- where you have enough spare memory that
>> your OS can cache the ENTIRE index.  The ideal setup is usually not
>> required for good performance.
>> Thanks,
>> Shawn

View raw message