We noticed few process instances getting stuck in the gateway step in production which is very strange as it should be a straight forward transitioning out. This issue is just for few instances and other instances went thru just fine. We want to know the reason for the same. From the logs we noticed " Unable to connect to Broker. Starting to poll" messages during that time and not sure if this has anything to do with this issue.

Any help or guidance in this will be appreciated.

Hi Balachandar,

as long as you dont restart IS while Broker is unavailable there should be no issues with these messages.
IS will buffer the messages locally.

Are there any messages in the server.log or error log related “Unable to publish” or “Rejecting Document for Trigger …”?

Check why IS is not able to connect to the broker.
Is this temporarily or permanentky?

Regards,
Holger

This is just a one time issue we noticed and for some reason before even we notice the issue the servers got restarted. All we see are the “Unable to connect to Broker. Starting to poll” messages in our server log during that time.

Hi Balachandar,

in this case you can try to resubmit the process at the nearest possible point before the gateway.
Another option might be to just stop the process instance(s) when the rest of the processing is out of date now.

When Broker is unavailable and the IS gets restarted the locally buffered documents are lost.
In this case the transition documents, which contain the information for the process models which step willl be executed next.

Regards,
Holger

Hello,

Holger, if you set those documents as “guaranteed”, don’t they get written to disk before being sent to the broker?

Balachandar, what do you find on the broker logs at the time the IS said it couldn’t connect to the broker? Do you have the cpu and io being monitored on that machine?

It could be a network problem between both machines (if the IS and Broker are separated) or too much CPU and/or IO usage on the broker server which made it unresponsive.

Best Regards,

Unfortunately we couldn’t notice any issues in the broker server apart from what we noticed in the server logs “Unable to connect to Broker. Starting to poll”

Hi,

Any difference in process instances that went through and that didnt? was business payload different for the models that got stuck?
what type of gateway step is it?