I'm trying to see whether there are any max files or max files per process issues with
Mac OS X. But so far I haven't found anything. As I mentioned to Charles earlier
today, it looks like there were approx. 250 emails (500 files) in my inbox around the
time things went funny, so we're not talking big numbers.
Someone on the Mac Java Developer group suggested I check the output of
sysct -a | grep files
which returns:
kern.maxfiles = 12288
which sounds okay from what I can gather.
One odd thing is that, if I re-run restart JAMES it refuses to let me download my
email. So there is something in there that is causing it to fail.
Does JAMES store any permanent status info in a file - numbers of files waiting to
process, or anything like that? If it does and one of these is a crazy value, perhaps
that could throw things?
Stuart.
On Thursday, November 23, 2000, at 10:18 AM, Charles Benett wrote:
> Well, this is not the problem I thought it might be.
> Any other details?
> Has anybody else noticed this?
> Charles
>
> Stuart Roebuck wrote:
> >
> > Last night I encountered something which sounds a bit like what has been described
>below,
> however, in putting the logs together below I note the fault appeared to start
>during
> receipt of an email by SMTP. I'm running James 1.2 on MacOS X Beta (roughly Java
>1.2.2 +
> Hotspot).
> >
> > Everything was working fine for 24 hours (I installed Monday evening).
> > Last night I tried to download my email by POP3 and was refused login. I'll refer
>to this
> situation as 'the fault'.
> >
> > Here is all the information (relevant or irrelevant) that I have gathered:
> >
> > Logging in using the admin facility still appears to work after the 'the fault'.
>The
> 'help' command responds correctly, but 'listusers' doesn't work and hangs. If I
>restart
> James 'listusers' works again.
> >
> > Looking at the logs I notice a line in avalon.log which may coincide with 'the
>fault':
> >
> > 2000.11.22 09:40:07 193 (ERROR) from Avalon: .getBlock: Exception in block null
>init:
> null
> >
> > Looking at jamesfile.log there is a clear point when 'the fault' starts. Here is
>the trace
> just before and then into the errors that repeat many times a second at various
>intervals
> (perhaps coinciding with the arrival of mail by SMTP):
> >
> > 2000.11.22 12:45:59 266 (INFO) from Processor: ==== Begin processing mail
> Mail974853959111-2697 ====
> > 2000.11.22 12:45:59 281 (INFO) from Processor: Processed Mail974853959111-2697
> through transport
> > 2000.11.22 12:45:59 282 (INFO) from Processor: Result was ghost
> > 2000.11.22 12:45:59 350 (INFO) from Processor: ==== Removed from spool mail
> Mail974853959111-2697 ====
> > 2000.11.22 12:51:49 372 (ERROR) from Processor: Exception in JamesSpoolManager.run
> null
> > 2000.11.22 12:51:49 388 (ERROR) from Processor: Exception in JamesSpoolManager.run
> null
> >
> > Looking at netfile.log there appears to be a clear problem occuring with the
>number of
> open files. The first fault I can find occurs during receipt of an email by James
>(SMTP):
> >
> > 2000.11.22 12:51:48 788 (INFO) from SMTP: Command received: EHLO
> relaysmtp.digiweb.com
> > 2000.11.22 12:51:48 909 (INFO) from SMTP: Command received: MAIL
> From:<[EMAIL PROTECTED]>
> > 2000.11.22 12:51:49 028 (INFO) from SMTP: Command received: RCPT
> To:<[EMAIL PROTECTED]>
> > 2000.11.22 12:51:49 148 (INFO) from SMTP: Command received: DATA
> > 2000.11.22 12:51:49 344 (ERROR) from SMTP: Error processing message: Exception
> spooling message: Exception caught while storing Message Container:
> java.lang.RuntimeException: Exception caught while storing a stream:
> java.io.FileNotFoundException:
> ../var/mail/spool/4D61696C3937343835343330393332332D32363939.private.StreamStore
> (Too many open files)
> >
> > From this point on the log file appears to contain the full text of the email
>received.
> Presumably it has got out of sync with receiving commands and is taking every line
>of the
> email as an SMTP command. The 'too many open files' fault appears to reoccur for
>every email
> received.
> >
> > Then, having restarted James, there is what appears to be a related fault during
>an
> attempt to download email (by POP3). Here's an anonymised log of a POP3 connection
> exhibiting 'the fault':
> >
> > 2000.11.22 09:40:07 017 (INFO) from POP3: Connection from mailclient
> (192.168.xxx.xxx)
> > 2000.11.22 09:40:07 039 (INFO) from POP3: Command recieved: USER stuart.roebuck
> > 2000.11.22 09:40:07 045 (INFO) from POP3: Command recieved: PASS xxxx
> > 2000.11.22 09:40:07 100 (INFO) from POP3: Connection from mailclient
> (192.168.xxx.xxx)
> > 2000.11.22 09:40:07 116 (INFO) from POP3: Command recieved: USER postmaster
> > 2000.11.22 09:40:07 118 (INFO) from POP3: Command recieved: PASS xxxx
> > 2000.11.22 09:40:08 097 (INFO) from POP3: Command recieved: STAT
> > 2000.11.22 09:40:08 872 (INFO) from POP3: Command recieved: LIST
> > 2000.11.22 09:40:08 948 (INFO) from POP3: Command recieved: UIDL
> > 2000.11.22 09:40:09 148 (INFO) from POP3: Command recieved: QUIT
> > 2000.11.22 09:40:09 156 (INFO) from POP3: Connection closed
> > 2000.11.22 09:40:23 425 (ERROR) from POP3: Exception during connection from
> mailclient (192.168.xxx.xxx) : Exception caught while retrieving an object:
> java.io.FileNotFoundException:
>
>../var/mail/localinbox/stuart/roebuck/4D61696C3937343835303130383531332D32353335.private.PersistentStore
>
> (Too many open files)
> >
> > Stuart.
> >
> > > Charles I am using 1.2 and it is still causing the problem.
> > > It also occurs on ordinary email, it affects pop dnloads and the
> > > outgoing mail. The lack of <CRLF> on the end of a message with a not
> > > empty final line causes the system to hang (only tested with netscape NT
> > > client) when pop dnloading. I think this also may be contributing to the
> > > file delete problems with NT. When a pop dnload is cancelled due to it
> > > hanging, the file handle is still kept in the operation system
> > > preventing the file from been deleted.
> > >
> > > What appears to be the problem is that the CharTerminatedInputStream
> > > class is consuming the <CRLF>.<CRLF> at the end of the message. It needs
> > > to leave the the first <CRLF> or even the whole <CRLF>.<CRLF> at the
> > > expense of an extra . in the message. I below is a solution I am
> > > working on that does not consume the <CRLF>.<CRLF> I have not had a
> > > chance to test it yet, you may want to try it though.
> > >
> > > pb..
> > >
> > >
> > > <code filename=3D"CharTerminatedInputStream2.java">
> > >
> > > package org.apache.james.util;
> > -------------------------------------------------------------------------
> > Stuart Roebuck, BSc, MBA Tel.: 0131 228 4853 / Fax.: 0870 054 8322
> > Managing Director Alpha Numeric Pager: 07654 588898
> > ADOLOS http://www.adolos.com/
> >
> > ------------------------------------------------------------
> > To subscribe: [EMAIL PROTECTED]
> > To unsubscribe: [EMAIL PROTECTED]
> > Archives: <http://www.mail-archive.com/james%40list.working-dogs.com/>
> > Problems?: [EMAIL PROTECTED]
>
>
> ------------------------------------------------------------
> To subscribe: [EMAIL PROTECTED]
> To unsubscribe: [EMAIL PROTECTED]
> Archives: <http://www.mail-archive.com/james%40list.working-dogs.com/>
> Problems?: [EMAIL PROTECTED]
>
>
-------------------------------------------------------------------------
Stuart Roebuck [EMAIL PROTECTED]
Lead Developer Mac OS X, Java, XML, etc.
ADOLOS http://www.adolos.com/
------------------------------------------------------------
To subscribe: [EMAIL PROTECTED]
To unsubscribe: [EMAIL PROTECTED]
Archives: <http://www.mail-archive.com/james%40list.working-dogs.com/>
Problems?: [EMAIL PROTECTED]
