Hi Varun,
Thanks, I will try that and open the jira.
Btw, is it possible (and reasonable?) to create large segments during
bulk-indexing (without interleaved commits) and then merge them into fewer
(e.g. 40) segments when committing at the end? So far my feeling (without
logging enabled) is that
HI Nawab,
I don't see a way to turn it of directly from Solr. Can you please create a
Jira for this?
In your case do you see evidence why you need to turn it off? I'd suggest
turning on Lucene level info stream logging by enabling then in the
resources/log4j.properties file . Index the documents
varun:
How do you set auto-IOThrottle? Is there a way to turn it off? I don't see
a setting for it.
Nawab
On Fri, Aug 4, 2017 at 9:49 AM, Varun Thacker wrote:
> I just ran into this issue yesterday and a couple of times last week.
>
> The team I was working with was
I just ran into this issue yesterday and a couple of times last week.
The team I was working with was running Solr 5.4.1 so the auto-throttle
merge policy applies. We were using the default merge policy. When
infoStream was turned on for a dummy collection on the same box the output
for the
Hi,
It's documented in the Javadocs of ConcurrentMergeScheduler. It depends on the
number of CPUs (with some upper bound) and if the index is on SSD. Without SSD
it uses only one thread for merging.
Uwe
Am 2. August 2017 22:01:51 MESZ schrieb Nawab Zada Asad Iqbal
:
>Thanks
Thanks Uwe,
That worked actually. After running for 3 hours, I observed about 88% of
indexing rate as compared to 4.5.0 without any file descriptor issues. It
seems that I can probably do some tweaking to get same throughput as
before. I looked at the code and the default values for
Hi,
You should reset those settings back to defaults (remove the inner settings in
the factory). 30 merge threads will eat up all your file handles. In earlier
versions of Lucene, internal limitations in IndexWriter make it unlikely, that
you spawn too many threads, so 30 had no effect.
In
Thanks Jigar
I haven't tweaked ConcurrentMergeStrategy between 4.5.0 and 6.6. Here is
what I have:
30
30
On Mon, Jul 31, 2017 at 8:56 PM, Jigar Shah wrote:
> I faced such problem when I used nomergepolicy, and did some code to manual
> merging segments which had
Hi Erick,
I am committing once after all the indexing finishes. This is a batch
indexing scenario to prepare new servers or replicas, and we don't use this
config for everyday operations. While batch indexing, we also modify code
to skip IndexWriter.maybeApplyDeletes to optimize the commit.
I
I faced such problem when I used nomergepolicy, and did some code to manual
merging segments which had bug and I had same issue. Make sure you have
default AFAIR ConcurrentMergeStrategy enabled. And its is configured with
appropriate settings.
On Jul 31, 2017 11:21 PM, "Erick Erickson"
No, nothing's changed fundamentally. But you say:
"We have some batch indexing scripts, which
flood the solr servers with indexing requests (while keeping open-searcher
false)"
What is your commit interval? Regardless of whether openSearcher is false
or not, background merging continues apace
Hi,
I am upgrading from solr4.5 to solr6.6 and hitting this issue during
complete reindexing scenario. We have some batch indexing scripts, which
flood the solr servers with indexing requests (while keeping open-searcher
false) for many hours and then perform one commit. This used to work fine
12 matches
Mail list logo