tracing pathsend error 904 (201)

[Henrik Paludan-Mørk]

A number of programs in my installation have a bad habit of throwing 901 errors (pathsend error) with the 201 subcode.
Seems to mean something about a pathway server receiving your pathsend call, crashes dreadfully before fully coming alive. the program abends.

I am a bit confused about how to handle/diagnose such a fault.

Any ideas? I am on J06.20.00.
the called servers are on high-pin.

//henrik

[Gustavo Cavazos]

Have you exhausted all the usual things to do:
1- check saveabend
2- compile programs with saveabend on and symbols
3- run the porgram with a terminal attached to see other outputs
4- check EMS
5- Can you re-create the error at-will? if so debug manually...

901 error means pathmon name is wrong?

just a few ideas..

gcav

[wbreidbach]

Error 901 says "Invalid Pathmon name". Be sure to fill the correct name and the correct length which is very likely the problem. $ABC has a length of 4, if you specify the length as 5 you would get an error 901.

[wbreidbach]

As the headline says 904: 904 is definitely a programming error, in that case Gustavo's hints are pretty useful.
In addition: Define the serverclass with numstatic 1 and try to start the serverclass. Check the result with status <serverclass>,detail. If there is an error that would be the reason for the pathsend problem.

[Roberto Veldhoven]

It is indeed a bit confusing that you write 904 in the header and 901 in the text. It must be 904 to get this system-error 201 though.

You do not tell if this is a production system or test/development.
If it is production the executable should have been properly tested and you may want to check the configuration of the serverclass. E.g. when you specify a hometerm that does not exist the process may fail when it attempts to write to it. In a test-environment the same can happen of course.

[Henrik Paludan-Mørk]

I better supply more detail.

production system. spurious error. actual error number is 904 (not 901). does not seem to appear with any regularity. i do not know how to provoke the error as i do not understand the cause or potential sources of the error.

We are producing heaps of saveabends. we have to delete 10.000 ZZSA files every now and then. highly complex program.

this is why i am here. have tried dumping memory areas to log. can see no errors.

[wbreidbach]

Have a look at the PATHMON log, there might be some information. An abending process should send a message to its hometerminal so specifying a hometerminal that preserves such messages would be very useful.
We are using a process that writes all these messages to files in a round-robin mode.

[Dave]

I am a little confused about what these 10,000 ZZSA files represent. Are they from the program performing the Pathsend and then abending because it gets a 201 or the server abending at some point. If it is the client you can inspect a saveabend and see where it is sending then look to see why that is causing the 201. If they are from the server you should be able look at why it is dying.

[Henrik Paludan-Mørk]

Dave, it's the calling program. I can see all the data areas, and they all look pretty OK to me. I think i will place a bet on the pathmon process log.

[dave....@gmail.com]

So that tells me you can look to see what server it is sending to. Since that server is abending you should be able to find a saveabend for it and then debug that.

[Henrik Paludan-Mørk]

I know which server is being sent to, and i can see the calling server (client) making a ZZSA abend file afer the failure. I do not get any error other than the 904/201 and the called server does not produce a ZZSA file. This is why i am here asking. I dont seem to get much info at all. I will try to get something from the PATHMON log next time it occurs.

[Dave]

So it appears there might be an intermittent startup failure of the server. Seems like it could maybe me a resource problem that would probably be reported in EMS. Or maybe a linking issue which would be in the Pathmon log but also probably in EMS. Maybe you should keep a few of those servers always running to see if the issue disappears. If it does it might give you a direction to head.

[Randall]

Another random thought: Does your server run LOW-PIN? That is a very limited resource.

[Roberto Veldhoven]

When you write that "the called server does not produce a ZZSA file", I assume that you have checked that its executable has the SaveAbend flag set?

You may want to start a Measurement for the processes of the called server
ADD PROCESS <PROGRAMFILENAME-OF-CALLED-SERVER>
and after a few of these failures verify that new processes had indeed been started at the time of the failures (or not). That will at least give you some idea of what goes on: e.g. how many messages were received before it failed.