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; } } }

Mutiple OD config files sharing the same DD config

cliffhanger

Environment:
TeamSite 6.5
OD 6.0.2
SOLARIS

We have multiple opendeploy config files referencing the same DD config file. With Older version of OD this was no problem, but with OD 6.x we're getting error when two different opendeploy config file has DNR scripts that invokes the datadeploy using the following:

iwodcmd start dd_glf -k iwdd=DEPLOYMENT_NAME

and are run at the same time

Has anyone had similar issues? If so, what did you do to avoid this?

We're currently using symbolic links with different names for each deployment but all links pointing to the same DD config file:

iwodcmd start dd_glf_A -k iwdd=DEPLOYMENT_NAME
iwodcmd start dd_glf_B -k iwdd=DEPLOYMENT_NAME

where dd_glf_A and dd_glf_B both are soft links to dd_glf.

Does anyone see issues with this approach?

Find more posts tagged with

Comments

Migrateduser

Instead, try using the -inst flag which allows you to run multiple instances of the same OD definition. Say,

iwodcmd start dd_glf -k iwdd=DEPLOYMENT_NAME -inst OD1
iwodcmd start dd_glf -k iwdd=DEPLOYMENT_NAME -inst OD2

(for instance)

cliffhanger

Thanks Dave!!

Wondering now how did I forget all about the -inst Smiley Sad

Anyway, thanks a bunch!!

-cliff

Migrateduser

Not a problem -- watch your log files, though. Each variation of inst starts a new log file and even though you might have your max log file size relatively low, this can skyrocket quickly when you run a lot of these things which may quickly devour disk space.

jed

Even with "inst", I think the DD's will be serialized on the backend. This is generally not a big issue, but if you have many large deployments, this can get old fast.

To enable parallel deploys, set execProcess='true' in the odbase.xml file. I have had mixed results with this, so make sure you test before dropping that in production.

From the DTD:


     execProcess= Deployments run from the command line are always
                  added to a single queue. To have parallel deployments
                  to happen set this attribute to "yes"

Migrateduser

Good point, thanks. So, it depends where the bottleneck is (OD vs. DD) but it makes good sense to take both precautions.

Curious, where did you have problems with DD in this situation? I can't think of a database platform whose drivers wouldn't serialize transactions anyway, even if you were locking tables.

cliffhanger

My bottleneck was OD, not DD. I don't mind if the deployments are serialized, I'd rather prefer it that way.

But thanks for the info Jed, in case such scenario comes, where we want parallel DD deployments.

jed

Curious, where did you have problems with DD in this situation? I can't think of a database platform whose drivers wouldn't serialize transactions anyway, even if you were locking tables.

In some situations, we will have a "standalone" DD process hang which causes everything else to stack up. I was hoping that enabling parallel deployments would prevent this.

When we ran in parallel mode during multiple large deployments (1,000+ files), this seemed to cause an endless loop. I have not performed a proper investigation, so it may have been other issues at work. I am no DBA, but table/row/database locking may have contributed to our issues.