Problems with updatemanager hanging

Question

Hello,
I was just wondering if other people are having problems with updatemanager hanging when you install patches. Quit frequently we have problems where the updatemanger will get 75% done and just hang. I have even let it sit over the weekend on our test boxes and it does not proceed. You can click on cancel and the same thing nothing happens.
What is the recommend procedure for stopping updatemanger when this happens. I am assuming a kill -15, I was hoping for something more graceful with less risk.
Thanks,
Gary

Answer 1

Update Manager community:
First, thanks to David Holmes for his suggestion that this is a Java bug related to the PKCS11 library forking problem. This is, at the very least, consistent with my observations: when things hang, it appears to hang in the java process and never seems to get where you see the various pkgadd and pkginstall processes get started.
I've made the suggested re-ordering of security providers in /path_to_java/jre/lib/security/java.security and will be watching anxiously to see if this resolves things.
Mr. Forum Moderator, have you and the other folks in UpdateManager-cental evaluated Mr. Holmes observation that this is the cause of many of the updatemanager hangs problems?
While I'm waiting to see if the changes to java.security change things, here is the set of things that I've found that help me to deal with updatemanager.
Note: my experiences apply to Solaris 10 (now u5) x86. Specifically, I am running on 3 nearly identical X4100 M2 platforms. The hanging problem seems to be intermittent in that sometimes 2 machines will update just fine and the third one will hang and it is not always the same machine that hangs. Other times, all 3 will update without hanging. When it does hang, however, prstat shows that the pkgadd, pkginstall, and related programs and scripts that do the real work never get fired off. And, as others have observed, once updatemanager hangs in this fashion, the "Cancel" button doesn't help either.
So, when things hang, these are the things that I've found to help in allowing me to restart to see if things will proceed properly the next time.
1. ps -ef | grep update
That usually shows 2 processes related to updatemanager (one a shell script and one the actual java program) and two instances of smpatch with a list of patches.
2. kill those 4 PIDs found in step one.
3. rm /var/sadm/spool/sunucLog/job_history.log.lck
If this *.lck file is still in place, updatemanager won't even check for new patches.
4. rm /tmp/.ai*
There is sometimes a file that starts with /tmp/.ai_and_a_very_long_string .... I don't know what it does, but it seems to need to be removed for another updatemanager run to be started.
5. ps -ef | grep ccr
This usually shows 2 or more cc-ccr processes. I believe that these also need to be killed (sometimes with a kill -9) and you probably need to check to make sure that init doesn't restart these.
I certainly don't claim to be an expert on these matters .... but, at the very least, I've found that these steps help me to deal with a hanging updatemanager.
Of course, I'm hopeful that David Holmes' observations related to the PCKS11 library bug turns out to be the identification of the underlying problem .... and I thank him again for his Java debugging skills and digging through the bug database.
Thanks,
John