[
http://jira.dspace.org/jira/browse/DS-628?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Tim Donohue updated DS-628:
---------------------------
Fix Version/s: 1.7.0
Assignee: Claudia Jürgen
This issue was discussed in the DSpace Developers Meeting on Sept 29 2010.
Claudia, we decided to assign this to you for a small update before committing
it to Trunk (so it can be released in 1.7).
The small update is that we feel the default timeout should be bumped up
slightly to 200 milliseconds. That way, hopefully most people will see fewer
solr timeouts in their logs in out-of-the-box DSpace. We thought about
bumping it up to 2 seconds by default, but there were some concerns that this
may not be the best for all DSpace instances. See full discussion for more
info.
[20:17] <tdonohue> Make the timeout for the extended resolver dnslookup
configurable : http://jira.dspace.org/jira/browse/DS-628
[20:18] <gilbertsonk> +1
[20:19] <PeterDietz> +1, I've personally needed to do this (hard coded) during
stats import
[20:19] <tdonohue> +1, this seems like a logical fix, and relatively small
change.
[20:19] <mhwood> +1
[20:20] * mdiggory (~mdigg...@ip72-199-217-116.sd.sd.cox.net) has joined
#duraspace
[20:20] <tdonohue> anyone want to volunteer to do a quick test and commit it
for 1.7?
[20:20] <kshepherd> do we want a "no lookup, addresses only" configuration
option too?
[20:20] <kshepherd> oh, it's reverse
[20:20] <PeterDietz> The patch looks good, would claudia want it assigned to
her?
[20:21] <tdonohue> sure, we can assign to Claudia and just ask her to commit it
[20:21] <tdonohue> DS-628 : assign to Claudia -- approved for committing to
Trunk
[20:22] <tdonohue> Question for DS-628 actually -- do we want to change the
default here? if 20ms is problematic on most systems, why not change it?
[20:23] <tdonohue> (noticed the patch has the config set to 20ms in dspace.cfg)
[20:24] <PeterDietz> Its been a few months since my SOLR days, I'm not sure I
know best
[20:24] <mhwood> That makes sense. 200ms?
[20:25] <tdonohue> it looks like Claudia is suggesting it should normallly be
set to 2000-5000ms (2-5 secs).
[20:26] <mdiggory> The Solr client uses apache HTTPClient and farms out async
threads from a pool for each request. I suspect there will be a threshold where
the pool will get exhausted on significant loads. So there will be a trade off,
do not set this too high.
[20:26] <mhwood> What happens at high request rates if many are timing out?
[20:26] <mhwood> Wow, asnwered before asked!
[20:26] <tdonohue> mdiggory -- any suggestions as to what it should be set to
then? does 20ms make the most sense, or 200ms (I'd just like to determine what
is a reasonable default for "most people")
[20:27] <mhwood> We may have to just pick something and see how it works in the
field, unless there are data already.
[20:28] <kshepherd> how it works during an intense crawl will be quite
different to how it works in a quiet time, so it might not be immediately
apparent, either
[20:28] <tdonohue> ok -- we'll leave at 20ms then for now, I guess. I wasn't
sure if anyone here had more experience with this and already had to "bump it
up" (similar to Claudia)
[20:28] <mdiggory> bump it up to 200, if we see continued cases, then bump it
up further
[20:29] <tdonohue> mdiggory -- ok. For DS-628 assign to claudia and ask her to
bump to 200ms for default. Commit to trunk
> Make the timeout for the extended resolver dnslookup configurable
> -----------------------------------------------------------------
>
> Key: DS-628
> URL: http://jira.dspace.org/jira/browse/DS-628
> Project: DSpace 1.x
> Issue Type: Improvement
> Components: DSpace API
> Affects Versions: 1.7.0
> Reporter: Claudia Jürgen
> Assignee: Claudia Jürgen
> Priority: Minor
> Fix For: 1.7.0
>
> Attachments: DnsLookup.java.patch
>
>
> This patch enables the configuration of the extended resolver's time out for
> the dns lookup.
> A new configuration parameter solr.resolver.timeout (sorry for yet another
> config param) is introduced and used to set the timout in DnsLookup.java.
> For backward compatibility the time out defaults to 20 milliseconds.
> OS time outs vary between 2 to 5 seconds, from what I've learned so far.
> We did face an increasing number of solr errors (200- 900/day) due to
> timeout, with the timeout being set to 20 milliseconds. Setting it to the
> system defaults (5 seconds) got rid of most of these. Nowadays an average of
> 2/day do occur.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://jira.dspace.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
------------------------------------------------------------------------------
Start uncovering the many advantages of virtual appliances
and start using them to simplify application deployment and
accelerate your shift to cloud computing.
http://p.sf.net/sfu/novell-sfdev2dev
_______________________________________________
Dspace-devel mailing list
Dspace-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dspace-devel
