Bugzilla – Bug 4790
stdout RP gets null
Last modified: 2008-02-05 15:24:30
You need to log in before you can comment on or make changes to this bug.
Hello Peter, We have an intermittent problem when obtaining the stdout RP value of a ws-gram job. We make 2 calls to obtain the value in case if fails. One call is made after the job becomes active but sometimes the active notification is not delivered or sent. The 2nd call is made after the 'Done' notification is received and before the ws-gram job object is destroyed. Sometimes it happens that we do not receive the active notification and that the RP value is null when probing after receiving the Done notification. the job i'm executing is a shell script which takes little time to execute. I'm using GT4.0.3 Thierry.
Created an attachment (id=1081) [details] Gemlca log showing null RP I've attached a part of a GEMLCA log. Please look at ServiceThread-6. A job is submitted. There is no Active notification delivered. It goes sraight into Cleanup and Done. The RP is checked right after the job is done (as we did not get an active notification) but the RP is null. I can usually reproduce this when submitting multiple (4 to 8) jobs at the same time to ws-gram.
Reassigning to current GRAM developer to close/fix as appropriate.
We have not seen this behavior with recent versions of WS GRAM (4.0.5 and 4.0.6) We've had 500 concurrent globusrun-ws clients which fetch the stdout RP. There have been many improvements since 4.0.3 and apparently this has been fixed.