Re: epoll on squid

If I have time today, I will try and setup an environment similar to
your own on one of my test boxes. I did have this working on my desktop
under Redhat 8.0, although I have whacked the box since then so I don't
remember the build environment on that. FYI, the test box is currently
running a Gentoo i686 build with linux 2.5.52, gcc 3.2, and
glibc-2.2.5. There is a glibc-2.3.1 ebuild script in the portage tree
which I can try building later this afternoon.

I am going to try and test the squid instance in production today to see
how it performs.

On Wed, 2002-12-18 at 05:29, Reuben Farrelly wrote:
> Hi David,
>
> At 05:25 PM 17/12/2002 -0500, you wrote:
>
> >I tried everything from scratch, by pulling the branch out of CVS and
> >building it myself, but everything worked fine for me still even with
> >the new linux kernel 2.5.52 and the newest tarball from squid-3.0-HEAD.
>
> Hrm, I wonder if we are running into glibc problems...I am running an unmodified 2.5.52 kernel too...glibc is the only main difference I guess..
>
> >> 2002/12/18 00:51:21| Loaded Icons.
> >> 2002/12/18 00:51:21| commSetSelect: cpoll_ctl(,EPOLL_CTL_MOD,,) failed on
> >> fd=14!: (0) Success
> >
> >The epoll_ctl(,EPOLL_CTL_DEL,,) errors are probably related to the fact
> >that it didn't call epoll_ctl(,EPOLL_CTL_ADD,,) previously, but instead
> >failed through. Perhaps if we work around it for now by doing this:
> >
> >Change this:
> >if (epoll_ctl(kdpfd, EPOLL_CTL_MOD, fd, &ev) < 0) {
> > if(errno == ENOENT) {
> >To this:
> >if (epoll_ctl(kdpfd, EPOLL_CTL_MOD, fd, &ev) < 0) {
> > if((errno == ENOENT) || (errno == 0)
>
> I had to make some small changes ie: (and yes, word wrap is not great in Eudora)
>
> ----------------------------------------
> --- comm_epoll.cc-orig 2002-12-18 20:45:08.000000000 +1100
> +++ comm_epoll.cc 2002-12-18 20:48:40.000000000 +1100
> @@ -144,7 +144,7 @@
> ev.events = events;
> if(events) {
> if (epoll_ctl(kdpfd, EPOLL_CTL_MOD, fd, &ev) < 0) {
> - if(errno == ENOENT) {
> + if((errno == ENOENT) || (errno == 0)) {
> debug(5,4) ("commSetSelect: epoll_ctl(,EPOLL_CTL_MOD,,) failed on fd=%d: entry does not exist\n",fd);
> if (epoll_ctl(kdpfd, EPOLL_CTL_ADD, fd, &ev) < 0)
> debug(5,1) ("commSetSelect: cpoll_ctl(,EPOLL_CTL_ADD,,) failed on fd=%d!: %s\n",fd,xstrerror());
> ------------------------------------------
>
>
>
> >If that gets it working for you, let me know, and I will add it into the
> >CVS branch.
>
> It gets past the initial fail point, now I get this, and again I can't actually request anything:
>
> 2002/12/18 21:07:27| Starting Squid Cache version 3.0-DEVEL for i686-pc-linux-gnu...
> 2002/12/18 21:07:27| Process ID 18904
> 2002/12/18 21:07:27| With 1024 file descriptors available
> 2002/12/18 21:07:27| DNS Socket created at 0.0.0.0, port 32916, FD 6
> 2002/12/18 21:07:27| Adding nameserver 203.29.67.170 from /etc/resolv.conf
> 2002/12/18 21:07:27| Adding nameserver 203.29.67.23 from /etc/resolv.conf
> 2002/12/18 21:07:27| Unlinkd pipe opened on FD 11
> 2002/12/18 21:07:27| Swap maxSize 1638400 KB, estimated 182044 objects
> 2002/12/18 21:07:27| Target number of buckets: 9102
> 2002/12/18 21:07:27| Using 16384 Store buckets
> 2002/12/18 21:07:27| Max Mem size: 4096 KB
> 2002/12/18 21:07:27| Max Swap size: 1638400 KB
> 2002/12/18 21:07:27| Store logging disabled
> 2002/12/18 21:07:27| Rebuilding storage in /var/spool/squid-1 (CLEAN)
> 2002/12/18 21:07:27| Rebuilding storage in /var/spool/squid-2 (CLEAN)
> 2002/12/18 21:07:27| Using Least Load store dir selection
> 2002/12/18 21:07:27| Set Current Directory to /var/spool/squid
> 2002/12/18 21:07:27| Loaded Icons.
> 2002/12/18 21:07:27| Accepting HTTP connections at 203.29.67.170, port 3128, FD 14.
> 2002/12/18 21:07:27| Accepting ICP messages at 0.0.0.0, port 3130, FD 15.
> 2002/12/18 21:07:27| Accepting SNMP messages on port 3401, FD 16.
> 2002/12/18 21:07:27| Accepting WCCP messages on port 2048, FD 17.
> 2002/12/18 21:07:27| Ready to serve requests.
> 2002/12/18 21:07:28| Store rebuilding is 6.1% complete
> 2002/12/18 21:07:32| Done reading /var/spool/squid-1 swaplog (66642 entries)
> 2002/12/18 21:07:32| Done reading /var/spool/squid-2 swaplog (80319 entries)
> 2002/12/18 21:07:32| Finished rebuilding storage from disk.
> 2002/12/18 21:07:32| 146961 Entries scanned
> 2002/12/18 21:07:32| 0 Invalid entries.
> 2002/12/18 21:07:32| 0 With invalid flags.
> 2002/12/18 21:07:32| 146961 Objects loaded.
> 2002/12/18 21:07:32| 0 Objects expired.
> 2002/12/18 21:07:32| 0 Objects cancelled.
> 2002/12/18 21:07:32| 0 Duplicate URLs purged.
> 2002/12/18 21:07:32| 0 Swapfile clashes avoided.
> 2002/12/18 21:07:32| Took 5.4 seconds (27190.0 objects/sec).
> 2002/12/18 21:07:32| Beginning Validation Procedure
> 2002/12/18 21:07:32| Completed Validation Procedure
> 2002/12/18 21:07:32| Validated 146961 Entries
> 2002/12/18 21:07:32| store_swap_size = 1263096k
> 2002/12/18 21:07:33| storeLateRelease: released 0 objects
> 2002/12/18 21:08:09| commSetSelect: cpoll_ctl(,EPOLL_CTL_MOD,,) failed on fd=13!: (115) Operation now in progress
> 2002/12/18 21:08:23| commSetSelect: cpoll_ctl(,EPOLL_CTL_MOD,,) failed on fd=18!: (11) Resource temporarily unavailable
>
>
> >Also, what build environment and configure options are you using?
>
> Redhat 8, with some updates from rawhide, ie glibc-2.3.1-6, and gcc-3.2.1 I'm pulling all my squid source from CVS too and have all up to date trees checked out. I'm not patching my 2.5.52 kernel, it's just a stock build..
>
>
> Reuben
>
> -------------------------------------------------------------
> Reuben Farrelly West Ryde, NSW 2114, Australia

-- 
David Nicklay
Location: CNN Center - SE0811A
Office: 404-827-2698	Cell: 404-545-6218