André Warnier wrote:
Sam, because there is quite a bit of traffic on this issue, and because
the exact description seems to change a bit over time (e.g.
compilation-only or not), could you remake a summary of what you are
sure of so far ?
1. Initial requests to ANY JSP page on a site, precompiled or not, or a compiled servlet, take an inordinate amount of
time to generate a response, often several minutes, as long as 20 minutes or so after the server is restarted.
2. After a response is finally generated from a given JSP, that JSP always
responds quickly to subsequent requests.
3. The four virtual hosts start serving pages at the same time, after a Tomcat restart, as long as 20 minutes after
Tomcat started up.
4. jstack showed that the delay during a small JSP request was during a call to java.io.File.exists(). It is still not
known what the particular file was.
5. strace reveals that thousands and thousands of stat64, access and lstat64 calls are made to JAR files and classes
during the initial "stall phase." I've posted a couple short examples from my test6.jar requests.
Without totally excluding it for instance, I would surmise that it is
unlikely to be a disk-related problem, because copying the complete
webapp somewhere else does not change the issue.
Yes, and I have done so several times.
About the File.exists() issue and what it is checking, isn't a tool like
"lsof" a possible way to find out ? It can show all I/O like resources
used by a given process, be they sockets or files.
I'm looking at lsof now. Haven't seen anything remarkable so far, but I'll
keep at it a bit.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org
