Date: Sat, 01 Mar 2003 08:20:56 -0700
From: "Peregrine M. McGehee"
Subject: Re: 2.5m obslog 20030228 (MJD 52699)
Pete Newman wrote:
> Spectrograph focus was checked and adjusted during the afternoon.  See
> "Spectro Focus" section for details.  Watcher reported an MCP:tpm error we
> have not seen before:
>
>         ca (INVALID: 2003-03-01 01:39:51Z)
>
Pete,
This indicates that the channel access (CA) query between the
client (IOP tpmhandler on sdsshost) and the servers (TPM, TPM2)
timed out. The default CA timeout is 30 seconds - so something
must have been very wrong - this usually means the server
was down.
However, both TPMs claim uptimes (at this writing) of ~140 hours
so this wasn't a reboot or hang. Other options are:
	o high network traffic
	o high CPU load on sdsshost
	o high CPU load on TPM or TPM2 - the CA communications runs
	at a lower priority than the device drivers and the real-time
	database.
A standard EPICS tool (that we don't have installed) is the iocStats
package that reports IOC (vxWorks box) resource usage - which could
get archived. Something to add for the March BT?
Regards,
	Peregrine
--
Peregrine M. McGehee	http://astro.nmsu.edu/~peregrin
Department of Astronomy
New Mexico State University, Las Cruces, NM