Site stops responding with tcpip protocol

[Elisha Gould]

Hi,

I’ve noticed that with some of our sites that use multiple outbound threads with the tcpip protocol that there is an occational race condition causing the site to freeze when one/both of the threads disconnect for some time due to the remote system dropping the connection.

The site has a single inbound thread, runs through some xlates, then sends to both outbound threads. The site is also slow to start if the threads are opening and data sent takes a long time to get to the outbound queue (6 messages takes around 1 minute).

If I change the protocol to pdl-tcpip and write an appropriate pdl to process the messages, everything runs normally.

has anyone else notice this issue?

We are running Cloverleaf 5.8.5.4 on RHEL 5.7.

Topic

[Jeff Dinsmore]

Elisha,

We are experiencing a similar phenomenon with one process that delivers to 21 localhost connections using the tcpip protocol.

This only happens during the midnight to 01:00 hour, when we’re receiving daily outpatient auto-discharges (A03).  There are a few hundred during a this hour or so.  It does not occur over the weekend when outpatient volume is much lower.

The effect seems to be that, because of the backups in this one process – whatever the cause may be – it’s  unable to ACK messages for some time and the sender times out.  The sender drops it’s connection and tries to connect.  When does, it sends more messages and process repeats.

Did you find the root cause of this problem?

Has anyone else seen/solved this problem while still using the tcpip protocol?

Jeff Dinsmore
Chesapeake Regional Healthcare

[Author]

Replies