nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

scalability issues w/ OD??

jed

We have a support note open, but hopefully someone else has dealt with this before...

OD 6.0.1 on solaris 9

We have noticed some scalability issues with OD. We there are ~15-20 concurrent deployments running, we start to see connection errors. We get the following in the log file:

>>>
ENG: 2005-01-28 16:27:00 PST GMT-08:00 Thread-5702 Connecting to host=fooooo.com port=30014

ENG: 2005-01-28 16:27:00 PST GMT-08:00 Thread-5702 Sender Side timeout = 0

ENG: 2005-01-28 16:27:04 PST GMT-08:00 Thread-5702 ***ERROR - Job ID=m2324Deployment=wft_filelist_prod.167951
Base: Exceeded all 3 retries to get a connection with the target OpenDeploy host.
Failed creating socket.
LIB: 2005-01-28 16:27:04 LOGFILE=[/app/suncom/opendeploy/OpenDeployNG/log/src.wft_filelist_prod.167951.ALL_CONTENT.barrrrrrr.log]
<<<

When things settle down, a redeploy will go through without issues. Anyone have any advice besides "run less jobs at one time"?

Thanks

--
Jed Michnowicz
jedm@sun.com
Content Management Engineer
Sun Microsystems

Find more posts tagged with

Comments

Migrateduser

Could be maxing out on deployment legs.

You might try increasing the number of FDs and making sure the NSCD is enabled, per the OD 6.0.1 Admin Guide, pp. 165-168.

Todd Scallan
Director of Product Management
Interwoven
t: 408-530-7167
e: tscallan@interwoven.com

jed

Thanks Todd. I'll check out it.

--
Jed Michnowicz
jedm@sun.com
Content Management Engineer
Sun Microsystems

jed

Does the use of SSL increase the legs in use? We have been able to reproduce this in our dev enviro, and the problem goes away when we disable SSL. We are doing the following:

- File list deployment with one file of 300k
- SSL enabled
- 1 DNR that runs the following command: sleep 120 (this is so deploys get stacked up)
- invoke deployments using a script with a 2 second pause between each deploy

after ~35 deploys, we start to see the connection failures

If we disable SSL, we have been able to run 80+ OD's concurrently.

Any ideas?
Thanks
~jed

--
Jed Michnowicz
jedm@sun.com
Content Management Engineer
Sun Microsystems

Edited by jed on 02/01/05 04:18 PM (server time).

Migrateduser

I don't think SSL should be a factor in terms of leg limits.

You should have Tech Support recreate the problem in house. They can then get Engineering to analyze the issue and respond with a workaround or fix.

Todd Scallan
Director of Product Management
Interwoven
t: 408-530-7167
e: tscallan@interwoven.com

Freedom

Has the issue ever been resolved? I am running into the same error message now.

Thanks

jed

Short story:
Fixed in the most recent od602 patch.

Long story:
Our issue was a result of using SSL deploys. WIthout SSL, the problem would not occur. It was pretty easy to reproduce this by doing something along the lines of:
o deploy one file using SSL
o add a sleep statement as a DNR for ~2 minutes so deploys stack up
o loop a bunch of deploys concurrently until you get an error

To get around this, we used multiple instances of the OD base server and created a scheduler find the instance that was under the lowest load. While this worked, it was an administrative nightmare trying to keep configs in sync (odbase, odnodes, conf/*). Using ant, xslt's, mappings of all the instances and key/value substitutions, we were able to ease the config file pain, but it is something I NEVER want to deal with again. Anyway, install the patch...

--
Jed Michnowicz
jedm@sun.com
Content Management Engineering
Sun Microsystems