Re: alh CA Disconnects

[Marty Kraimer <mrk@aps.anl.gov>]

Paul Sichta wrote:
> 
> I have been trying to use alh and am getting sporadic (every 2-3 hours) CA
> disconnect of groups of records.  The alh config file has records in two
> IOC's.  When a disconnect occurs,  all records attached to an IOC disconnect
> and then reconnect the next second.  This happens to both IOC but not at the
> same time.
> 
> The system is lightly loaded.  The TMO variable is 30 seconds and the BEACON
> is 15 seconds.  medm clients and other CA clients on the host do not
> disconnect.  This behavior been the case with the latest alh(1.1.2) and a
> former version.

Are you absolutely certain that no other clients see this. The reason I ask is
that connection management is handled by CA and not the clients. Thus if all
your clients are build against the same version of CA they should have the same
connection management behavior.
 
> The error message I get is shown below (one message for each
> mass-disconnect, not one for each record).  The transient PV alarm states
> are  COMM and INVALID,  and then back to NO_ALARM again the next second.
> 
> Using R3.13.0beta11, mv177 IOC, solaris Host.  "Private" network.  Any
> clues?


Perhaps you are having something happen in your network of which you are not
aware. Do you have a metwork sniffer? Did you pay attention to the discussion
last Dec by Garrett D. Rinehart IPNS here at ANL? It had a lot of valuable
information.

Another possibility. Are your IOCs heavily loaded for some period of time. Since
CA is lowest priority it will not be able to do anything if something else grabs
the cpu. If this happens for a long time the clients will think they lost
communication.

Marty Kraimer