[squid-users] httpAccept: FD 51: accept failure: (130) Software caused connection abort.
Hello all, I have been working for ING bank in Belgium. I am responsible for the system administration of the Squid environnement to let our internal employees surf on the Internet. We get 15000 employees, all surfing through our 8 solaris servers. You can find below characteristics for all these servers : SUNW,Sun-Fire-V210 SunOS 5.8 Generic_117000-03 sun4u sparc 2048 Mb Memory 2 x Gbps bge Network Interfaces 2 x internal disks ( soft mirror except for the cache partition ) 2 X 20Go Squid Cache ( one on each disk ). If I look at the statistics, we currently get 1753764 cached objects on one of my Squid servers and such 60 HTTP Requests per second. We also already optimized our Solaris machines by setting up two caches separately defined on two disks, all of them mounted with next options /dev/dsk/c0t0d0s5 /dev/rdsk/c0t0d0s5 /ING/SQUID/var/cache0 ufs 2 yes logging,noatime /dev/dsk/c0t1d0s5 /dev/rdsk/c0t1d0s5 /ING/SQUID/var/cache1 ufs 2 yes logging,noatime We also defined next IPC parameters on each host set shmsys:shminfo_shmseg=16 set shmsys:shminfo_shmmni=32 set shmsys:shminfo_shmmax=2097152 set msgsys:msginfo_msgmni=40 set msgsys:msginfo_msgmax=2048 set msgsys:msginfo_msgmnb=8192 set msgsys:msginfo_msgssz=64 set msgsys:msginfo_msgtql=2048 But we get a problem. Our HTTP traffic is increasing a bit every day and I noticed today we get lots of these messages in cache.log. 2007/04/02 07:59:27| comm_accept: FD 51: (130) Software caused connection abort 2007/04/02 07:59:27| httpAccept: FD 51: accept failure: (130) Software caused connection abort 2007/04/02 08:16:27| comm_accept: FD 51: (130) Software caused connection abort 2007/04/02 08:16:27| httpAccept: FD 51: accept failure: (130) Software caused connection abort 2007/04/02 08:19:50| comm_accept: FD 51: (130) Software caused connection abort 2007/04/02 08:19:50| httpAccept: FD 51: accept failure: (130) Software caused connection abort 2007/04/02 08:22:48| comm_accept: FD 51: (130) Software caused connection abort 2007/04/02 08:22:48| httpAccept: FD 51: accept failure: (130) Software caused connection abort 2007/04/02 08:22:50| comm_accept: FD 51: (130) Software caused connection abort 2007/04/02 08:22:50| httpAccept: FD 51: accept failure: (130) Software caused connection abort I restarted process squid this morning but this did not solve the problem. After some search on the net, it seems we get some network buffers problems. It could be we can solve the problem by increasing some values in /etc/system but I do not know which one I can increase. You can find below some results from these next statements : cat cache.log ulimit -a squidclient mgr:info /usr/sbin/sysdef -i I would appreciate your help because all of these machines are running in our production environment and we can get in trouble if I do not solve it. Regards. *** 2007/04/02 07:55:48| Starting Squid Cache version 2.5.STABLE12 for sparc-sun-solaris2.8... 2007/04/02 07:55:48| Process ID 24393 2007/04/02 07:55:48| With 1024 file descriptors available 2007/04/02 07:55:48| Performing DNS Tests... 2007/04/02 07:55:48| Successful DNS name lookup tests... 2007/04/02 07:55:48| DNS Socket created at 0.0.0.0, port 61044, FD 6 2007/04/02 07:55:48| Adding nameserver 10.66.122.32 from squid.conf 2007/04/02 07:55:48| Adding nameserver 10.66.67.3 from squid.conf 2007/04/02 07:55:51| Referer logging is disabled. 2007/04/02 07:55:51| Unlinkd pipe opened on FD 47 2007/04/02 07:55:51| Swap maxSize 3072 KB, estimated 2363076 objects 2007/04/02 07:55:51| Target number of buckets: 118153 2007/04/02 07:55:51| Using 131072 Store buckets 2007/04/02 07:55:51| Max Mem size: 409600 KB 2007/04/02 07:55:51| Max Swap size: 3072 KB 2007/04/02 07:55:51| Local cache digest enabled; rebuild/rewrite every 3600/3600 sec 2007/04/02 07:55:51| Store logging disabled 2007/04/02 07:55:51| Rebuilding storage in /ING/SQUID/var/cache0 (CLEAN) 2007/04/02 07:55:51| Rebuilding storage in /ING/SQUID/var/cache1 (CLEAN) 2007/04/02 07:55:51| Using Least Load store dir selection 2007/04/02 07:55:51| Current Directory is /ING/SQUID 2007/04/02 07:55:51| Loaded Icons. 2007/04/02 07:55:51| Accepting HTTP connections at 10.66.184.64, port 8080, FD 51. 2007/04/02 07:55:51| Accepting HTTP connections at 0.0.0.0, port 8080, FD 53. 2007/04/02 07:55:51| Accepting ICP messages at 0.0.0.0, port 3130, FD 54. 2007/04/02 07:55:51| Accepting HTCP messages on port 4827, FD 55. 2007/04/02 07:55:51| Accepting SNMP messages on port 3401, FD 56. 2007/04/02 07:55:51| WCCP Disabled. 2007/04/02 07:55:51| Configuring Parent 10.66.9.233/8080/0 2007/04/02 07:55:51| Configuring Parent 10.66.9.232/8080/0 2007/04/02 07:55:51| Configuring Parent 10.66.9.235/8080/0 2007/04/02 07:55:51| Configuring Parent 10.66.9.234/8080/0 2007/04/02 07:55:51| Configuring Parent 10.66.17.36/7001/0 2007/04/02 07:
Re: [squid-users] httpAccept: FD 51: accept failure: (130) Software caused connection abort.
[EMAIL PROTECTED] wrote: Hello all, I have been working for ING bank in Belgium. I am responsible for the system administration of the Squid environnement to let our internal employees surf on the Internet. SNIP But we get a problem. Our HTTP traffic is increasing a bit every day and I noticed today we get lots of these messages in cache.log. 2007/04/02 07:59:27| comm_accept: FD 51: (130) Software caused connection abort 2007/04/02 07:59:27| httpAccept: FD 51: accept failure: (130) Software caused connection abort 2007/04/02 08:16:27| comm_accept: FD 51: (130) Software caused connection abort 2007/04/02 08:16:27| httpAccept: FD 51: accept failure: (130) Software caused connection abort 2007/04/02 08:19:50| comm_accept: FD 51: (130) Software caused connection abort 2007/04/02 08:19:50| httpAccept: FD 51: accept failure: (130) Software caused connection abort 2007/04/02 08:22:48| comm_accept: FD 51: (130) Software caused connection abort 2007/04/02 08:22:48| httpAccept: FD 51: accept failure: (130) Software caused connection abort 2007/04/02 08:22:50| comm_accept: FD 51: (130) Software caused connection abort 2007/04/02 08:22:50| httpAccept: FD 51: accept failure: (130) Software caused connection abort http://www.squid-cache.org/mail-archive/squid-users/200401/0239.html Chris
Re: [squid-users] httpAccept: FD 51: accept failure: (130) Software caused connection abort.
Thanks a lot for the info Chris. On Tue, April 3, 2007 01:18, Chris Robertson wrote: [EMAIL PROTECTED] wrote: > Hello all, > > > I have been working for ING bank in Belgium. I am responsible for the > system administration of the Squid environnement to let our internal > employees surf on the Internet. > SNIP > But we get a problem. Our HTTP traffic is increasing a bit every day and > I noticed today we get lots of these messages in cache.log. > > > 2007/04/02 07:59:27| comm_accept: FD 51: (130) Software caused > connection abort 2007/04/02 07:59:27| httpAccept: FD 51: accept failure: > (130) > Software caused connection abort > 2007/04/02 08:16:27| comm_accept: FD 51: (130) Software caused > connection abort 2007/04/02 08:16:27| httpAccept: FD 51: accept failure: > (130) > Software caused connection abort > 2007/04/02 08:19:50| comm_accept: FD 51: (130) Software caused > connection abort 2007/04/02 08:19:50| httpAccept: FD 51: accept failure: > (130) > Software caused connection abort > 2007/04/02 08:22:48| comm_accept: FD 51: (130) Software caused > connection abort 2007/04/02 08:22:48| httpAccept: FD 51: accept failure: > (130) > Software caused connection abort > 2007/04/02 08:22:50| comm_accept: FD 51: (130) Software caused > connection abort 2007/04/02 08:22:50| httpAccept: FD 51: accept failure: > (130) > Software caused connection abort > > http://www.squid-cache.org/mail-archive/squid-users/200401/0239.html Chris
Re: [squid-users] httpAccept: FD 51: accept failure: (130) Software caused connection abort.
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On Mon, 2 Apr 2007 09:57:43 +0200 <[EMAIL PROTECTED]> wrote: > > Hello all, > > I have been working for ING bank in Belgium. I am responsible for the > system administration of the Squid environnement to let our internal > employees surf on the Internet. > > We get 15000 employees, all surfing through our 8 solaris servers. You > can find below characteristics for all these servers : > >SUNW,Sun-Fire-V210 >SunOS 5.8 Generic_117000-03 sun4u sparc >2048 Mb Memory >2 x Gbps bge Network Interfaces >2 x internal disks ( soft mirror except for the cache partition ) >2 X 20Go Squid Cache ( one on each disk ). > > If I look at the statistics, we currently get 1753764 cached objects on > one of my Squid servers and such 60 HTTP Requests per second. We also > already optimized our Solaris machines by setting up two caches > separately defined on two disks, all of them mounted with next options > >/dev/dsk/c0t0d0s5 /dev/rdsk/c0t0d0s5 /ING/SQUID/var/cache0 > ufs 2 yes logging,noatime >/dev/dsk/c0t1d0s5 /dev/rdsk/c0t1d0s5 /ING/SQUID/var/cache1 > ufs 2 yes logging,noatime > > We also defined next IPC parameters on each host > >set shmsys:shminfo_shmseg=16 >set shmsys:shminfo_shmmni=32 >set shmsys:shminfo_shmmax=2097152 >set msgsys:msginfo_msgmni=40 >set msgsys:msginfo_msgmax=2048 >set msgsys:msginfo_msgmnb=8192 >set msgsys:msginfo_msgssz=64 >set msgsys:msginfo_msgtql=2048 > > But we get a problem. Our HTTP traffic is increasing a bit every day and > I noticed today we get lots of these messages in cache.log. > >2007/04/02 07:59:27| comm_accept: FD 51: (130) Software caused > connection abort >2007/04/02 07:59:27| httpAccept: FD 51: accept failure: (130) > Software caused connection abort >2007/04/02 08:16:27| comm_accept: FD 51: (130) Software caused > connection abort >2007/04/02 08:16:27| httpAccept: FD 51: accept failure: (130) > Software caused connection abort >2007/04/02 08:19:50| comm_accept: FD 51: (130) Software caused > connection abort >2007/04/02 08:19:50| httpAccept: FD 51: accept failure: (130) > Software caused connection abort >2007/04/02 08:22:48| comm_accept: FD 51: (130) Software caused > connection abort >2007/04/02 08:22:48| httpAccept: FD 51: accept failure: (130) > Software caused connection abort >2007/04/02 08:22:50| comm_accept: FD 51: (130) Software caused > connection abort >2007/04/02 08:22:50| httpAccept: FD 51: accept failure: (130) > Software caused connection abort > > I restarted process squid this morning but this did not solve the > problem. After some search on the net, it seems we get some network > buffers problems. It could be we can solve the problem by increasing > some values in /etc/system but I do not know which one I can increase. > You can find below some results from these next statements : > > cat cache.log > ulimit -a > squidclient mgr:info > /usr/sbin/sysdef -i > > I would appreciate your help because all of these machines are running > in our production environment and we can get in trouble if I do not > solve it. > > Regards. > > *** > > 2007/04/02 07:55:48| Starting Squid Cache version 2.5.STABLE12 for > sparc-sun-solaris2.8... > 2007/04/02 07:55:48| Process ID 24393 > 2007/04/02 07:55:48| With 1024 file descriptors available > 2007/04/02 07:55:48| Performing DNS Tests... Hi Vincent, I think your Squid server is starting with only 1024 file descriptors as shown by your logs above. Since you mentioned that you have a user base of 15000 and traffic utilization is increasing, 1024 file descriptors may not be enough. This could result in poor performance. However, since I have no experience regarding Solaris OS, your Squid may not suffer from running out of file descriptors. > 2007/04/02 07:55:48| Successful DNS name lookup tests... > 2007/04/02 07:55:48| DNS Socket created at 0.0.0.0, port 61044, FD 6 > 2007/04/02 07:55:48| Adding nameserver 10.66.122.32 from squid.conf > 2007/04/02 07:55:48| Adding nameserver 10.66.67.3 from squid.conf > 2007/04/02 07:55:51| Referer logging is disabled. > 2007/04/02 07:55:51| Unlinkd pipe opened on FD 47 > 2007/04/02 07:55:51| Swap maxSize 3072 KB, estimated 2363076 objects > 2007/04/02 07:55:51| Target number of buckets: 118153 > 2007/04/02 07:55:51| Using 131072 Store buckets > 2007/04/02 07:55:51| Max Mem size: 409600 KB > 2007/04/02 07:55:51| Max Swap size: 3072 KB > 2007/04/02 07:55:51| Local cache digest enabled; rebuild/rewrite every > 3600/3600 sec > 2007/04/02 07:55:51| Store logging disabled > 2007/04/02 07:55:51| Rebuilding storage in /ING/SQUID/var/cache0 (CLEAN) > 2007/04/02 07:55:51| Rebuilding storage in /ING/SQUID/var/cache1 (CLEAN) > 2007/04/02 07:55:51| Using Least Load store dir selection > 2007/04/02 07:55:51| Current Directory is /ING/SQUID > 2007/04/