Bug#402603: tomcat5.5: blocks on startup until log pipe is read

2007-02-26 Thread Adrian Bridgett
.. and of course, delete catalina.out file (and the creation of it in the init.d script) - it's annoying when you do grep ... * and it hangs! Adrian -- Adrian Bridgett - [EMAIL PROTECTED] GPG key available on public key servers -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of

Bug#402603: tomcat5.5: blocks on startup until log pipe is read

2007-02-26 Thread Marcus Better
.. and of course, delete catalina.out file (and the creation of it in the init.d script) - it's annoying when you do grep ... * and it hangs! Yes, that pipe is going away as soon as I get my syslogging code in jsvc to a working state... pgpg0CoHEvjEh.pgp Description: PGP signature

Bug#402603: tomcat5.5: blocks on startup until log pipe is read

2007-02-16 Thread Adrian Bridgett
I've fixed this locally in two ways: a) use cronolog and alter init.d (see attached diff) Pro: simple Con: end up with two logs b) using log4j Pro: catalina.log file has predictable name for log analysis Con: more complicated for b) you need to delete the -outfile and -errfile line from the

Bug#402603: tomcat5.5: blocks on startup until log pipe is read

2006-12-12 Thread Marcus Better
Ah, so the situation is the result of your local configuration which disables rotatelogs? No, it's in 5.5.20-4 in the archive. This might lower the importance of the bug (it's currently considered release critical). I think we should release 5.5.20-2 which is in testing. It has no RC

Bug#402603: tomcat5.5: blocks on startup until log pipe is read

2006-12-12 Thread Loïc Minier
On Tue, Dec 12, 2006, Marcus Better wrote: It's usually to diagnose startup problems. I propose that we add a debug mode, disabled by default, in which we send stdout and stderr to a regular file, or even the named pipe, and tell the user to read it manually. In normal operation we

Bug#402603: tomcat5.5: blocks on startup until log pipe is read

2006-12-12 Thread Marcus Better
But you need to bootstrap it somehow, and during the bootstrap, there's usually a simple logger which logs to stdout. I retract my previous proposal, and propose instead to send the initial messages to the syslog. That would be logical for a daemon. I found this blog post on the

Bug#402603: tomcat5.5: blocks on startup until log pipe is read

2006-12-12 Thread Loïc Minier
On Tue, Dec 12, 2006, Marcus Better wrote: I retract my previous proposal, and propose instead to send the initial messages to the syslog. That would be logical for a daemon. So, something like: start_daemon | while read line; do mylog $line; done with: mylog() { logger -p ... $* }

Bug#402603: tomcat5.5: blocks on startup until log pipe is read

2006-12-12 Thread Marcus Better
Here's another possibility, adding syslog support to jsvc: https://issues.apache.org/jira/browse/DAEMON-80 pgpuYWiGmZWXW.pgp Description: PGP signature

Bug#402603: tomcat5.5: blocks on startup until log pipe is read

2006-12-12 Thread Loïc Minier
On Tue, Dec 12, 2006, Marcus Better wrote: Here's another possibility, adding syslog support to jsvc: https://issues.apache.org/jira/browse/DAEMON-80 (Thinking out loud) Is it Debian specific to depend on jsvc to start tomcat? IIUC, jsvc has native libraries, it's an important design

Bug#402603: tomcat5.5: blocks on startup until log pipe is read

2006-12-12 Thread Marcus Better
Is it Debian specific to depend on jsvc to start tomcat? It's actually the method recommended by upstream: http://tomcat.apache.org/tomcat-5.5-doc/setup.html It's true that it requires a native app (not a library), but it seems to be an improvement over the previous method. For instance we

Bug#402603: tomcat5.5: blocks on startup until log pipe is read

2006-12-12 Thread Loïc Minier
On Tue, Dec 12, 2006, Marcus Better wrote: It's actually the method recommended by upstream Oh, perfect then. It's true that it requires a native app (not a library), but it seems to be an improvement over the previous method. For instance we get a clean shutdown procedure, which

Bug#402603: tomcat5.5: blocks on startup until log pipe is read

2006-12-12 Thread Marcus Better
(I don't recall running into that, nor that two JVMs thing, This was mostly an issue on an overloaded system, I think. Not sure what was going on, but the init script would never succeed in stopping Tomcat until the timeout elapsed. The second JVM invocation is the java call in catalina.sh

Bug#402603: tomcat5.5: blocks on startup until log pipe is read

2006-12-11 Thread Marcus Better
Package: tomcat5.5 Version: 5.5.20-4 Severity: serious tomcat5.5 (sometimes) blocks on startup since its standard output is redirected to the named pipe /var/log/tomcat5.5/catalina.out, and this blocks until the pipe is opened for reading. Thus tomcat5.5 will not startup until the after pipe is

Bug#402603: tomcat5.5: blocks on startup until log pipe is read

2006-12-11 Thread Loïc Minier
On Mon, Dec 11, 2006, Marcus Better wrote: tomcat5.5 (sometimes) blocks on startup since its standard output is redirected to the named pipe /var/log/tomcat5.5/catalina.out, and this blocks until the pipe is opened for reading. Thus tomcat5.5 will not startup until the after pipe is read, even

Bug#402603: tomcat5.5: blocks on startup until log pipe is read

2006-12-11 Thread Marcus Better
tag 402603 help thanks (Sounds close to #350131 and #270248.) Yes, except that I cut out the rotatelogs stuff from the init script because it looked ugly. That's why the process is blocking now. What should we do about it? I don't like the idea of having an extra rotatelogs process to keep

Bug#402603: tomcat5.5: blocks on startup until log pipe is read

2006-12-11 Thread Loïc Minier
On Mon, Dec 11, 2006, Marcus Better wrote: Yes, except that I cut out the rotatelogs stuff from the init script because it looked ugly. That's why the process is blocking now. What should we do about it? Ah, so the situation is the result of your local configuration which disables