> > > Indeed. Have you truss'ed the squid process to find out what it is doing?
> >
> > Yes I have. Here is a sample truss output, in case you could figure something out
> > from it. To me it looks like a really heavy polling compared to a busy production
> > machine with lots of requests ...
>
> Mmmhhh .... I have seen this behaviour when experimenting with
> libmallocs and compile options ...
> My squid uses open() instead of open64().
> What compiler did you use; can you try different ones ?
Currently I'm using gcc 2.8.1, but I can certainly use for example gcc 2.7.2.3
or even acc (Sun Commercial one). Today, I inserted the following lines into the
/etc/system file:
set maxusers=256
set rlim_fd_max=1024
set rlim_fd_cur=512
Let's see if there is any difference. -DUSE_POLL didn't seem to solve the problem ...
> > Another very interesting warning message I got by using command `squid -k shutdown`.
> >
> >
> > 1998/03/13 09:10:03| Preparing for shutdown after 3989 connections
> > 1998/03/13 09:10:03| Waiting 1 seconds for active connections to finish
> > 1998/03/13 09:10:17| FD 15 Closing HTTP connection
> > 1998/03/13 09:10:17| FD 16 Closing ICP connection
> > 1998/03/13 09:10:28| comm_select: select failure: (9) Bad file number
> > 1998/03/13 09:10:28| examine_select: Examining open file descriptors...
> > 1998/03/13 09:10:28| FD 4: (9) Bad file number
> > 1998/03/13 09:10:28| WARNING: FD 4 has handlers, but it's invalid.
> > 1998/03/13 09:10:28| FD 4 is a Socket called ''
> > 1998/03/13 09:10:28| lifetm:0 tmout:0 read:1e290 write:1c528
> > 1998/03/13 09:10:28| Select loop Error. Retry 1
> > 1998/03/13 09:10:28| Shutting down...
> > 1998/03/13 09:10:28| storeWriteCleanLog: Starting...
> > 1998/03/13 09:10:29| Finished. Wrote 353 lines.
> > 1998/03/13 09:10:29| Took 1 seconds ( 353.0 lines/sec).
> > CPU Usage: user 7550 sys 39
> > Maximum Resident Size: 0 KB
> > Page faults with physical i/o: 4559
>
> I've seen those errors w/ 1.2b, you are using stock 1.1.20?
Yes, this is plain Squid 1.1.20 + retry patch. I have tried without the retry
patch as well without any change in Squid's behaviour.
//Tomi Orava
Tomi Orava | The views and opinions expressed in this article
Oy LM Ericsson Ab | are those of the poster and not his employer.
LMF/ZT/Unix Administration
phone: +358 9 2992604
Received on Mon Mar 16 1998 - 02:14:48 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:39:23 MST