XGKS colliding with Oracle?
Date: 5 Aug 92 17:09:20 GMT
Message-ID: <51037_at_seismo.CSS.GOV>
Our X application infrequently causes orasrv on the database server to die. The environment:
Client machine: SUN 4, OS 4.1.1
Database hosts: DEC RISC, ULTRIX V4.0 and ULTRIX V4.2
Fronted GUI : X11R5
XGKS, the GKS library for the X Window System developed
at the University of Illinois
Oracle : rdbms 6.0.33.2.3
pro*c 1.3.18.1.1
orasrv 1.2.7.2.3
-------------------------
Scenario #1: orasrv hangs
The client crashes while establishing a connection to the database (in at least one case because the user control-C'd the application at startup).
The app got the attention of orasrv, and orasrv is now waiting for another packet from the dead process. orasrv waits forever doing network reads, accumulating massive time, and blocking all further connection requests.
Scenario #2: orasrv dies
We set a 30 second timeout so orasrv won't wait forever ('tcpctl timeout 30'). The timeout is handled, but now orasrv dies. From the orasrv.log:
DEBUG: waiting for connection request... DEBUG: remote network address is 192.33.134.23. Connection request from tyr.css.gov at 4-AUG-92 15:39:06
ERROR: timed out reading network string! <==== timeoutDEBUG: waiting for connection request... DEBUG: remote network address is 192.33.134.23. Connection request from tyr.css.gov at 4-AUG-92 15:39:36
ERROR: unable to read network string (54) <==== read failure DEBUG: waiting for connection request... <==== orasrv zapped
- end of orasrv.log, no orasrv process is running --
GKS Collision?
We thought the problem might be GKS asynch io colliding with Oracle signals; its malloc and free routines block *all* signals so they are uninterruptable. Since Oracle is signal sensitive, this sounded bad. We changed it to only block SIG_IO.
And we thought everything was fine until yesterday. :-)
We haven't been able reproduce the problem. We have run the app through malloc_verify and Purify. I have left orasrv in debug mode. I will set a UNIX trace on orasrv to see if we can catch its final gasp.
I'm stumped and am hoping that somebody will recognize this problem.
-jean
+-----------------------------------------------------------------------+Received on Wed Aug 05 1992 - 19:09:20 CEST
| Jean Anderson, DBA email: jean_at_seismo.css.gov |
| SAIC Open Systems Division, MS A2-F or: jean_at_esosun.css.gov |
| 10210 Campus Point Drive phone: (619)458-2727 |
| San Diego, CA 92121 fax: (619)458-4993 |
+-----------------------------------------------------------------------+
| std_disclaimer("mine"); |
+-----------------------------------------------------------------------+