SWA + Search

[savant]

Hi,

in the meantime I've scoured the forums for solutions, omscan'd and omtidy'd everything, managed to fix an inconsistency when running sxmkindex -r 00:00:00 (now finishes vs. crashing) - but:

Still no joy in using search either from SWA or Outlook. Using the api call directly gives an empty page.

i.e.
http://mail.server/api/user@domain/search?q=term

using sxsearch --direct=term "User Name" gives tons of results.

Neither show up in scalix-sis-search.log.

In trying to track down problems, I've edited the log4j.properties file for tomcat/webapps/api/WEB-INF/classes/log4j.properties and set "DEBUG" or "INFO" instead of "FATAL" for each API activity. I even changed the "additivity" for search to see if it would show up anywhere when I made that request (and yes, I restarted scalix-tomcat after changing settings).

Well, it didn't - the files "scalix-api-activity.log" and "scalix-swa-activity.log" (which I also modified the logger settings for) remain at 0 bytes no matter what I do.

If you have a better way to track down what's happening, please let me know - I've tried pretty much everything I found digging through the forums and release notes.

Regards,
Simon.

Code: Select all

[root@klmcom logs]# ls -lart
insgesamt 4052
-rw-rw-rw- 1 root root       0  1. Okt 01:29 manager.2007-10-01.log
-rw-rw-rw- 1 root root       0  1. Okt 01:29 localhost.2007-10-01.log
-rw-rw-rw- 1 root root       0  1. Okt 01:29 host-manager.2007-10-01.log
-rw-rw-rw- 1 root root       0  1. Okt 01:29 admin.2007-10-01.log
-rw-rw-rw- 1 root root       0  1. Okt 01:29 scalix-platform.log
-rw-rw-rw- 1 root root       0  1. Okt 01:29 scalix-api-activity.log
-rw-rw-rw- 1 root root       0  1. Okt 01:29 scalix-swa-remote-client.log
-rw-rw-rw- 1 root root       0  1. Okt 01:29 scalix-swa-activity.log
-rw-rw-rw- 1 root root       0  1. Okt 01:29 scalix-mobile.log
-rw-rw-rw- 1 root root   88455  1. Okt 12:17 catalina.2007-10-01.log
-rw-r--r-- 1 root root   46475  2. Okt 11:37 scalix-sis-search.log.old
-rw-rw-rw- 1 root root  371463  2. Okt 13:37 scalix-swa.log
-rw-rw-rw- 1 root root       0  2. Okt 13:55 manager.2007-10-02.log
-rw-rw-rw- 1 root root       0  2. Okt 13:55 localhost.2007-10-02.log
-rw-rw-rw- 1 root root       0  2. Okt 13:55 host-manager.2007-10-02.log
-rw-rw-rw- 1 root root       0  2. Okt 13:55 admin.2007-10-02.log
-rw-rw-rw- 1 root root  107431  2. Okt 19:52 scalix-api.log
-rw-rw-rw- 1 root root   31801  2. Okt 22:48 catalina.2007-10-02.log
-rw-rw-rw- 1 root root       0  3. Okt 02:09 manager.2007-10-03.log
-rw-rw-rw- 1 root root       0  3. Okt 02:09 localhost.2007-10-03.log
-rw-rw-rw- 1 root root       0  3. Okt 02:09 host-manager.2007-10-03.log
-rw-rw-rw- 1 root root       0  3. Okt 02:09 admin.2007-10-03.log
drwxr-xr-x 7 root root    4096  3. Okt 02:43 ..
-rw-rw-rw- 1 root root  121263  3. Okt 02:44 scalix-sis-search.log
-rw-rw-rw- 1 root root  690943  3. Okt 02:44 scalix-sis-indexer.log
-rw-r--r-- 1 root root  157943  3. Okt 02:44 catalina.out
-rw-rw-rw- 1 root root    8537  3. Okt 02:44 catalina.2007-10-03.log
drwxr-xr-x 2 root root    4096  3. Okt 02:45 .
-rw-rw-rw- 1 root root 1555198  3. Okt 03:03 scalix-res.log
-rw-rw-rw- 1 root root  907916  3. Okt 03:03 scalix-caa.log

[/code]

[jaime.pinto]

We just finished an upgrade from 11.1 to 11.2 on RHEL5, and just as I was afraid of, the problem transcended the upgrade. It's still there.

I followed the wiki instructions on rebuilding the search indexes for all users, to no use.

To the Scalix Support Team, we need some escalation on resolving this issue. This forum shows that the number of companies and users being affected is increasing rapidly. Please provide a strategy to help us post information that will help nail down this bug, and establish a prompt fix. We have an open support call as well.

[chris]

Hey Jaime,

if the instructions about restarting the indexes didn't help, then your support case is the exactly right way to go.

Chris

[savant]

Hi,

one more observation on the search issue - SWA search is still not working, neither is it working using the 11.2GA connector (server is 11.2 as well).

So in toto - after upgrade from 10.0.x (fully working on all fronts), following all the advice about rebuilding indexes and MIME pre-generation as well as a few other things found on the forums, the only case where search works is

using an outdated connector? SWA/current connector/API - all no go.

Sorry to sound a bit increduluous, but I've been on the receiving end of a whole lot of complaints about this - no closer to figuring out anything.

So yesterday I tried changing locale (as per some notes added on a - marked resolved a few versions ago - bug.
Didn't work then for whatever reason, so I discarded that solution.
Out of desperation, today I did the same again. To my (pleasant, I might add) surprise, after again setting system default locale to en_US.UTF-8, search magically works (API + SWA).

So - here's the moral of the story - PLEASE MAKE THIS MORE VISIBLE!

It took a monumental amount of digging around and trial and error (and good thing VMWare snapshots exist) to figure out this - thought resolved - issue.

Please include it - again, with MUCH better visibility - somewhere in the installation docs. Seems to happen with upgrades from 10.x, mostly.

So - what fixed this for me:

Code: Select all

/etc/init.d/scalix-tomcat stop ; omshut
change system locale to en_US.UTF-8 (/etc/sysconfig/i18n on redhat/fedora)
logout and re-login (to make the locale change active in your shell or it won't work)
/etc/init.d/scalix-tomcat start ; omrc

Just setting it in tomcat's startup script didn't work for me, it had to be the default system locale.

The ulimit / max.search.results changes to be able to cleanly index large message stores could also be helpful. They sure made my index rebuild finish vs. dying prematurely.

All in all - thanks for a great product, but please do keep working on the documentation

Regards,
Simon.

[jaime.pinto]

Ours has always been en_US.UTF-8, and the ulimit is unlimited.Still no deal.

[savant]

Jaime,

ulimit cannot be unlimited unless you didn't use "ulimit -n" (maximum open file descriptors) but just "ulimit" (maximum size of files written by the shell and its children). The number you need to adjust is the open file descriptors though ("ulimit -n <new limit>").

Okay, let me put the real procedure and all that I did beforehand -

ulimit -n 80000
(insanely high but oh well, 71K to be indexed)

edit sis.properties and swa.properties to adjust max search results/indexing behaviour, all to the same values beyond the max item count.

In sis.properties:
index.max.queue.size
search.max.results

In swa.properties:
swa.search.maxresults

omscan'd and omtidyallu'd the whole server (nice 3-hour break there, glad our deployment isn't larger).

Re-Index cache - fixing any errors that still may crop up - my mailstore did have some inconsistencies that made "sxmkindex -r 00:00:00" fail every time. Even after those were fixed, it would still fail until I reset the above ulimit/parameters.

So after all that, all of a sudden it worked.

BTW Jaime, do you get search results from the shell using sxsearch? I always got those, but they were malformed (comma instead of dot), so the locale change fixed that. If not, you probably have a different problem.

Regards,
Simon.

[jaime.pinto]

Hey savant

We followed your steps as well, but still no luck:
1) omscan -a -A -f -l /tmp/omscan.out
Took some time but everything looks fine

2) ulimit -n 256000
sxmkindex -r 00:00:00
Total messages processed: 231804

3) Increased in sis.properties:
index.max.queue.size=300000
search.max.results=300000

4) tried the command line search again
sxsearch -d pattern <some user>
no output

5) checked these logs
tail tomcat/logs/scalix-sis-search.log
2007-10-06 10:04:02,686 INFO [main] [SISConfig.init:45] Starting Scalix Search and Index Service (SIS) version 11.2.0.52
2007-10-06 10:04:02,753 INFO [main] [SISConfig.load:31] SIS initialized and ready for searching

tail tomcat/logs/scalix-sis-indexer.log
2007-10-06 10:04:02,717 INFO [main] [SISConfig.init:46] Starting Scalix Search and Index Service (SIS) version 11.2.0.52
2007-10-06 10:04:02,734 INFO [main] [SISConfig.loadConfigFromFile:131] SIS Indexer configuration loaded from file /var/opt/scalix/m2/sis/sis.properties
2007-10-06 10:04:02,738 INFO [main] [SISConfig.init:89] Index root is /var/opt/scalix/m2/indexes
2007-10-06 10:04:02,754 INFO [main] [SISConfig.load:32] SIS initialized and ready for indexing

Since it's being so difficult for support at Scalix to figure out (and others in the forum), it must be something that in their "normal" logic they would never do, therefore they can't even consider as a possible cause for trouble. I figured that we should be looking for something *very unique* we may have done to our system, and in nature very unusual.

On the other hand, we did find something very basic and significant that should help: the command-line search tool doesn't work. It would be reasonable to expect that once this is fixed the whole problem should get resolved.

Anybody can formulate *any* conjectures? We'll investigate and try.

[tantony]

@jaime.pinto
did you and the scalix support team find a solution for your problem and would you share the solution with us?

Thomas

[jaime.pinto]

Yes we did. This morning we gave them remote access to our system and eventually they figured out what was wrong. I'll ask them to either post the solution or let me know what they did.

[tantony]

Hi Jaime.

Is your problem solved? What was the solution?

[jaime.pinto]

I emailed scalix twice already asking what they did. No response yet,

[MailMan]

Hi Jamie,

Did you get a response yet? We are having the same problem.

Regards,

Ken

[MailMan]

I figured this out just now... You need to go to:

/var/opt/scalix/$instance/sis/sis.properties

And add the server's IP address to:

index.client.whitelist=
search.client.whitelist=

Then restart tomcat and then apache:

/etc/init.d/scalix-tomcat restart
apachectl restart

It should start working pretty much immediately after this.

Regards,

Ken

[savant]

Ah,

that's something that I did in between and forgot about - thanks for the reminder. Hope it helps tantony too.

Regards,
Simon.

[tantony]

Hi,

I already have the IP from the server and the loopback address in both values and nobody can get a search result in SWA
index.client.whitelist=192.168.255.4,127.0.0.1
search.client.whitelist=192.168.255.4,127.0.0.1

Thank you for your suggestion.