Anyway, my machine was down. Eventually got it going again (PSU issue), and started Jive. I have the init script in my /etc/init.d dir with appropriate links in the rc3.d and rc6.d dirs. However, although in the boot process Jive is reported to have started, it hasn’'t. Anyway, that might be a side issue.
The main problem is when I manually start Jive with the init script. On Mandrake the command is service start.
Anyway, Jive starts, in the info log it reports that ports 5222 & 5223 and conference have all started. That’'s about it. Normally if I can get the number of clients trying to connect below 20 then I get the log entry stating that Jive has finished starting (where it reports the version number). However, today the number of clients was very low (about 4 or 5), but Jive would not finish starting.
trying to connect with my client, in XML debug I see the server responds, requests authentication. My client sends authentication and that’'s about as far as you get.
I shutdown Jive and restarted it about 5 times before it eventually started. Each time I gave Jive about 10 or so minutes to start.
I know the hardware that Jive is running on is not the best, but once it is started I don’'t get any problems. There is about 200Mb RAM installed. The installation is via the RPM. There is no other Java installation on the machine. There are no JAVA env entries to put the installation of course.
Any ideas? I’‘ve got the server up now, and barring any unforseen events like power cuts and hardware failures I don’'t anticipate any other problems. Even my plugins are working and it quite happily handles around 80 or so clients.
i have only 64MB RAM And i serve about 70 people ( ~ 45 online). I’'m using Archlinux. We have electricity break down quite often (~ 7 today ). But it only affects Exodus clients wount reconnect normally. Jive itself is running ok, even after UPS fails to hold it online. After such break downs (when server goes off) there were only issues with linux itself. It was scanning disk, and refused to startup at once, only after second try. But Jive was not affected it seems. Maybe it was just luck.
I wonder if there is a problem with JM, if too many clients attemp to connect while the software is still booting. If for example the ServerSockets are setup before JM has completely finished startup it may get stuck with a resource issue…
As far as I can tell wroot doesn’'t have the same issue as his clients are not auto-connecting as soon as a socket is available.
Its just a guess, perhaps Matt or Gato could elaborate on the start up seqence.
One thing we improved in JM 2.1.4 (or 2.1.5) was to start the socket listeners as the last step in the server start up process. Previously, it was possible for clients to connect to the server while the server was being initialized.
However, the socket listeners are started up before the plugins are being loaded. So it is possible, for a couple of seconds, that some plugins are missing.
I think Jive is fantastic. It’'s just this startup problem. No probs this morning though :-).
I was ust a little weird yesterday as there were only a few clients trying to connect. I determine that by using netstat and counting the connections to 5222 quiet regularly, and the number didn’‘t move from 5. Normally it could be in excess of 60 if I try and restart the server during the middle of the day (rare). So I wasn’'t expecting any problems.
I have since found the following error in the nohup.out file in the bin dir:
I did a google for you exception, the only thing that came close was CharsetEncoder. A quick search of JM src indicates that this class is not being used directly by the software. It may be used indirectly via xml parsing…
Wonder if your java setup/OS setup is leading the JVM astray. It would be very hard to pin down, but I remember the JVM crashing a few times while running JM and it turned out to be that some additional OS patch levels were required for the particular JVM.
Wonder if your java setup/OS setup is leading the JVM
astray. It would be very hard to pin down, but I
remember the JVM crashing a few times while running
JM and it turned out to be that some additional OS
patch levels were required for the particular JVM.
I’‘m quite prepared for the OS to be at fault, and by no means pinning the blame at Jive’‘s door. It is quite an old install and not maintained well. I’'m planning to change the OS to a debian base to make updates oh so much easier.
Maybe I’'ll have to do the change sooner rather than later. Bring on 25 hour days