Re: CA Gateway problem

[Dirk Zimoch <dirk.zimoch@psi.ch>]

Hi

I'm running the gateway on Redhat Linux 9.0. Meanwhile, I can see the 
problem on 2 iocs, a 3.13.2 and a 3.13.9 system. The problem is 
independent of the client program and can be seen with 3.14.6 clients as 
well as with 3.13.2 clients.

In both cases the record is a bi, ZNAM="ERROR", ONAM="OK".

Reading the record of the 3.13.2 ioc always returns OK even when the 
record is in ERROR state. Trying to read the numerical value of the enum 
(with caget -n from R3.14.6 base) results in the following error message 
and returns 0:

CA.Client.Exception...............................................
    Warning: "No reasonable data conversion between client and server 
types"
    Context: "op=0, channel=X07MA-ID1-PLC:ERROR, type=DBR_TIME_SHORT, 
count=1, ctx="read failed""
    Source File: ../getCopy.cpp line 81
    Current Time: Fri Aug 06 2004 10:21:51.167738000
...................................................................
Read operation timed out: some PV data was not read.

In the gateway log is the following error message:

----------dump This=0x882a374---------
 dimension=0 app-type=16 Scalar
 prim-type=aitString(aitEnumString)
this=0x882a374 string=0x60497de8<OK>, length=2, buf length=3, type=Allocated
 ref-count=1
 total-bytes=47, data-size=11, element-count=1
 LocalDataFormat
--------------------------------------
filename="../../../../src/cas/generic/casStrmClient.cc" line number=607
server tool changed bounds on request - get notify with 
PV=X07MA-ID1-PLC:ERROR type=15 count=1

Aug 06 10:53:24 !!! Errlog message received (message is above)

Other logged errormessages are like this: (related to reboot, I think)
Jul 01 00:25:04 Warning: Virtual circuit unresponsive 172.19.150.24:5064
Jul 01 00:25:12 Warning: Virtual circuit disconnect 172.19.150.24:5064

and: (Can that confuse the gateway?)
Jul 05 12:39:37 gatePvData::monitor: ca_add_masked_array_event failed 
for X04SA-ES1-SMP:SCAN.D1DA:
 The requested data transfer is greater than available memory or 
EPICS_CA_MAX_ARRAY_BYTES


The record on the 3.13.9 system is alway shown as ERROR. The numerical 
value appears to be 14, which is nonsense. For both records the correct 
value is 1. Reading this record does not procuce any error messages.

When I create a new virtual PV by requesting the VAL fiels explicitely, 
everything is correct.

The number of connected channels is not so high, 2600 at the moment.

What is special about the records?
1) The values are enums
2) The values had not changed for weeks
3) medm screens are connected for weeks

Dirk

Lahey, Terri E. wrote:
> In the SLAC accelerator control system, we have seen a similar problem. We see that after awhile, a SEVERITY is not being updated (the value is constant, so we can't comment on that). The SEVERITY returned by the IOC changes, but the PV Gateway does not return the changed value. Restarting the CA Gateway fixes the problem, and then changes to SEVERITY are visible through the PV Gateway again.
>
> We are running on Solaris using PV gateway 2.0.0.0Beta11 compiled with R3.14.6. We see the same problem when we compile the PV gateway with the gnu or solaris compiler. Our IOC is running R3.13.6, and the UNIX client is DM (R3.13.2) and caget (R3.13.2).
>
> I think we see failures in SEVERITY returned by the PV gateway after the PV gateway connects to a very large number of PVs, eg. 17000 PVs. Is there some maximum number of PVs that we should support with the PV Gateway?
>
> Is this possibly a problem with mismatch of EPICS versions running on the CA client, CA gateway, and IOC?
>
> Dirk, Are you running on Solaris or Linux?  What version of EPICS is on your IOCs? And what version of EPICS are the clients that are connecting to the PV gateway?
>
>
> Terri Lahey
>
>
>
>
> > -----Original Message-----
> > From: Dirk Zimoch [mailto:dirk.zimoch@psi.ch] 
> > Sent: Thursday, August 05, 2004 5:07 AM
> > To: TECHTALK
> > Subject: CA Gateway problem
> >
> > Hi,
> >
> > I'm using CA gateway 2.0.0.0Beta8 compiled with R3.14.6 to read records 
> > from another subnet. Now I found that one enum has the wrong value 
> > though the gateway. It is a bi with the two states "OK" and "ERROR". The 
> > record was on ERROR for weeks, but when it changed to OK, I could not 
> > see the change through the gateway! The severity however is correct, it 
> > went from MAJOR to NO_ALARM. All other records on the same IOC seem to 
> > be OK. I have checkes that there is no second record with the same name. 
> > What's wrong?
> >
> > Dirk
> >
> > --
> > Dr. Dirk Zimoch
> > Swiss Light Source
> > Paul Scherrer Institut
> > Computing and Controls
> > phone +41 56 310 5182
> > fax   +41 56 310 4413


--
Dr. Dirk Zimoch
Swiss Light Source
Paul Scherrer Institut
Computing and Controls
phone +41 56 310 5182
fax   +41 56 310 4413