0 votes
629 views
in Interview Questions by (30.6k points)

1 Answer

0 votes
ago by (30.6k points)

A checklist that finds the cause of most "production is slow / broken" tickets in a clustered BPM / BAW environment, in the order that pays off:

  1. Scope it: one user, one app, one member, or everything? Compare members (hit each one directly through its port, bypassing the load balancer). Look at the time line: change windows, deployments (new snapshot?), batch jobs.
  2. Logs: SystemOut.log / SystemErr.log of every member around the time (grep for CWLLG, CWTBG, WSVR0605W hung threads, DSRA data source errors, CWSIx messaging, OutOfMemory), the IHS error_log and plug-in log for routing errors (503, no member available).
  3. Health of the engine: Process Admin > Event Manager > Monitor (queue depth, schedulers running?), Process Inspector for failed instances and stuck tokens (question 52), Instances failed list; Process Admin > Server Admin > Caches for cache statistics.
  4. Resources: JVM heap and GC (verbose GC or the PMI viewer), thread pools (WebContainer, EM / "BPM Process Server" thread pool), JDBC connection pool usage (Tivoli Performance Viewer > JDBC connection pools: waiting threads = pool too small or leaked connections), disk (logs, SIBus file store), CPU.
  5. Database: long-running statements (lsw_bpd_instance, lsw_task queries without indexes, PDW loads), lock waits, statistics out of date (runstats), table growth (instance cleanup overdue: run the instance deletion / BPMDeleteSnapshot housekeeping).
  6. Threads: three javacores 30 s apart on the slow member; look for threads stuck in the same place (Rhino loops, LDAP calls, remote web services, database waits).
  7. Network / security: LDAP response times (question 49), SSL handshakes to back ends, DNS.
  8. Capture for IBM: the BAW MustGather (logs, javacores, configuration, the Support Tools collectors) if you open a case.
# quick triage commands on a member
grep -E "CWLLG|CWTBG|WSVR0605W|OutOfMemory" SystemOut.log | tail -50
kill -3 <pid>; sleep 30; kill -3 <pid>; sleep 30; kill -3 <pid>       # javacores in the profile's logs dir
curl -sk -u admin:pw "https://member1:9443/rest/bpm/wle/v1/systems"     # engine answering on this member?

References

Related questions

0 votes
1 answer 817 views
0 votes
1 answer 1.1k views
0 votes
1 answer 3.6k views
0 votes
1 answer 3.0k views
asked Nov 8, 2018 by BPM Tips Admin (21.5k points)
0 votes
1 answer 1.8k views
0 votes
1 answer 1.1k views

723 questions

807 answers

98 comments

4.8k users

Join BPM Community Discord Channel

Welcome to BPM Tips Q&A, Community wiki/forum where you can ask questions and receive answers from other IBM BPM experts and members of the community. Users with 2000 points will automatically be promoted to expert level.
Created by Dosvak LLC
Our Youtube Channel
...