My experience with this kind of behavior is not Xindice specific, but Java on Solaris (ironically) is horrible. Not even alpha quality in my mind, the Java that does work on Solaris is so tweaked I'd rather write native code. There has actually been a lot of articles (and a lot of internal Sun memos) that talk about how shameful Java on Solaris is. I suspect your problems are related to the memory management & allocation issues on the Solaris JVMs. There was an article (I don't know from where, but if it is important to anyone, I could find it with some snooping) that mentioned there were over 100 severity one bugs open against the Solaris JVM just on malloc issues.
So what I am saying, check your JVM....
Basit
On Tuesday, June 17, 2003, at 03:54 PM, Kevin Ross wrote:
Ouch, sorry, I haven't run on solaris.
Has anyone else seen this?
-Kevin
-----Original Message----- From: David Hosier [mailto:[EMAIL PROTECTED] Sent: Tuesday, June 17, 2003 2:38 PM To: xindice-users@xml.apache.org Subject: RE: Strange Solaris behavior
Well, I tried that and the server cored at 324 documents into the process. I got the follow error: 1614 File Size Limit Exceeded - core dumped. And it still only updates in increments of 81. Any other suggestions? Thanks.
-----Original Message----- From: Kevin Ross [mailto:[EMAIL PROTECTED] Sent: Tuesday, June 17, 2003 2:15 PM To: xindice-users@xml.apache.org Subject: RE: Strange Solaris behavior
Not sure if this is related, but from my solaris days, I know we had to adjust the ulimit. From the readme, it suggests you do this on Linux:
ulimit -s 2048 for SDK 1.3.1
Give it a shot on solaris and let us know if it is of any help.
-Kevin Ross
-----Original Message----- From: David Hosier [mailto:[EMAIL PROTECTED] Sent: Tuesday, June 17, 2003 1:55 PM To: xindice-users@xml.apache.org Subject: Strange Solaris behavior
Greetings,
I am running Xindice 1.0 on Solaris 8 with JDK1.3.1, and I am having some strange behavior. We have an application that is loading 6500+ documents into the database. These are small documents; just object serializations using XMLEncoder. When we ran the application using a Xindice instance running on a Windows XP machine, the documents were all stored in less 1.5 minutes and I was able to get constant updates of the progress with "xindice ld -c /db/gateway". However, when we run Xindice on our Solaris 8 machines, the behavior is not the same. It takes about 1 minute to store 81 documents and when I issue the 'ld' command, I only get updates in increments of 81. For example, I will run the command about 1 minute into the process and get 81 documents listed. I will run the command again 20 seconds later and still get 81. I can keep doing this for a while with the same results. Then eventually I will get 162 documents, then 243, and so on in increments of 81 at lengthy intervals. This happened on two separate Solaris machines with the same specs.
This seems very strange since when run on Windows XP, I can get constantly updating figures with 'ld'. Plus it is agonizingly slower. Granted my Windows box is much faster than the Solaris machines, but I doubt it is 20+ times faster. Has anyone seen this type of behavior before? Is it normal for it to be that slow and update in groups of 81? Any suggestions would be great! Thanks for your time.
David Hosier Software Engineer Longview Software
