I recently (this weekend) attempted an upgrade to Java 2 on our web
servers, and came across a few problems someone might be able to help
with. Briefly:
Java 1.2 green threads won't start on one system: It just hangs taking
all processor time and memory. This is a Dual PII machine with SMP
running kernel 2.2.9 and I've reported this as a bug to Blackdown. It's
not my software: it happens even if I just run "java" on its own.
Java 1.2 *native* threads seems to work fine, and really fast too (no
JIT even) on the same machine, but software problems exist - those I
think are mine I've fixed, but something weird is happening wrt Postgres
6.5 - more detail later.
Java 1.2 green threads works perfectly on the develop machine (1 x
AMDK6/2)
Java 1.2 native threads works the same on this machine as on the SMP
one, same problems too.
I'm doing all this because I need a particular TimeZone bug fixed, which
was fixed in 1.1.8 or later, but there isn't a 1.1.8 for Linux.
OK, more detail:
Both systems are SuSE Linux 6.0 with some updates: Apache 1.3.6,
Postgres 6.5, Live Software's JRunPro 2.3.1 build 145. The develop
system has a single AMD K6/2 processor, is running the SuSE kernel
2.0.36, whereas the live server has twin Pentium IIs and is running a
home-cooked SMP 2.2.9 (with the ICMP DoS fix applied).
The problem with green threads on the SMP system is the big one at the
moment, as if that was fixed, I could deploy instantly.
The native thread problem I'm having seems to be down to software. The
JVM seems to work fine (odd seeing about 26 little Java processes
sitting there awaiting hits) and once I fixed my own code (mostly
involved removing statics and improving design so I didn't need them -
shouldn't have been there anyway) so it doesn't keep opening and never
closing Postgres connections, that seems to work fine - unless I do two
identical searches via the website at the same time, at which point the
two Postgres tasks suddenly take a long time, eat up all processor time,
then give up leaving Postgres in an unusable state. As the same problem
doesn't occur when running under Green threads, or when concurrent
searches are not looking for the same things, I don't know now whether
the problem is in my code, the JDBC driver, or the JVM.
--
Rachel
----------------------------------------------------------------------
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
