Re: [squid-users] Squid does not respond to TCP SYN when there are thousands of connection

On 15/02/2013 11:53 p.m., dahanhsi wrote:
> Hi Amos,
>
>
> 2013/2/15 Amos Jeffries <squid3_at_treenet.co.nz>:
>> On 15/02/2013 10:43 p.m., dahanhsi wrote:
>>> Thanks for your reply,
>>> provide more information below:
>>>
>>> 2013/2/15 Amos Jeffries <squid3_at_treenet.co.nz>:
>>>> On 15/02/2013 10:12 p.m., dahanhsi wrote:
>>>>> Hi all,
>>>>>
>>>>> I use squid as a reverse proxy, and make thousands of connection to the
>>>>> it.
>>>> Which version of Squid?
>>> I use Squid 2.7
>>
>>
>> Output of "squid -v" please.
> # squid -v
> Squid Cache: Version 2.7.STABLE9
> configure options: '--prefix=/usr' '--exec_prefix=/usr'
> '--bindir=/usr/sbin' '--sbindir=/usr/sbin'
> '--libexecdir=/usr/lib/squid' '--sysconfdir=/etc/squid'
> '--localstatedir=/var/spool/squid' '--datadir=/usr/share/squid'
> '--enable-async-io' '--with-pthreads'
> '--enable-storeio=ufs,aufs,coss,diskd,null' '--enable-linux-netfilter'
> '--enable-arp-acl' '--enable-epoll'
> '--enable-removal-policies=lru,heap' '--enable-snmp'
> '--enable-delay-pools' '--enable-htcp' '--enable-cache-digests'
> '--enable-underscores' '--enable-referer-log' '--enable-useragent-log'
> '--enable-auth=basic,digest,ntlm,negotiate'
> '--enable-negotiate-auth-helpers=squid_kerb_auth' '--enable-carp'
> '--enable-follow-x-forwarded-for' '--with-large-files'
> '--with-maxfd=65536' 'amd64-debian-linux'
> 'build_alias=amd64-debian-linux' 'host_alias=amd64-debian-linux'
> 'target_alias=amd64-debian-linux' 'CFLAGS=-Wall -g -O2'
> 'LDFLAGS=-Wl,-Bsymbolic-functions' 'CPPFLAGS='
>
>>
>>
>>>> What do you mean by "thousands of connections". 1's of tousands? 10's of
>>>> thousands? 100's of thousands?
>>> # netstat -nat|grep -i "80"|wc -l
>>> the result vary from 4651 to 9404
>>>
>>>>> There are one ten of all connections can not establish in TCP layer,
>>>>> because squid does not respond SYN-ACK to client's SYN packet. How can
>>>>> I solve it?
>>>>> Thanks
>>>>
>>>> Check ulimit settings for Squid?
>>>>
>>> # ulimit -a
>>> core file size (blocks, -c) 0
>>> data seg size (kbytes, -d) unlimited
>>> scheduling priority (-e) 20
>>> file size (blocks, -f) unlimited
>>> pending signals (-i) 16382
>>> max locked memory (kbytes, -l) 64
>>> max memory size (kbytes, -m) unlimited
>>> open files (-n) 655360
>>> pipe size (512 bytes, -p) 8
>>> POSIX message queues (bytes, -q) 819200
>>> real-time priority (-r) 0
>>> stack size (kbytes, -s) 8192
>>> cpu time (seconds, -t) unlimited
>>> max user processes (-u) unlimited
>>> virtual memory (kbytes, -v) unlimited
>>> file locks (-x) unlimited
>>>
>>>> Check your cache.log for messages about running out of filedescriptors?
>>> I set my limit.conf to:
>>> root soft nofile 655360
>>> root hard nofile 655360
>>
>> That does not answer the question. Squid may have been built or configured
>> with a limit of less than 655360 filedescriptors.
>> cache.log should tell you if Squid is reaching some limit like this.
> my cache.log:
> 2013/02/15 8:30:10| Starting Squid Cache version 2.7.STABLE9 for
> x86_64-debian-linux-gnu...
> 2013/02/15 8:30:10| Process ID 8136
> 2013/02/15 8:30:10| With 2048 file descriptors available
> 2013/02/15 8:30:10| Using epoll for the IO loop
> 2013/02/15 8:30:10| DNS Socket created at 0.0.0.0, port 6450, FD 6
> 2013/02/15 8:30:10| Adding nameserver 8.8.8.8 from /etc/resolv.conf
> 2013/02/15 8:30:10| User-Agent logging is disabled.
> 2013/02/15 8:30:10| Referer logging is disabled.
> 2013/02/15 8:30:10| logfileOpen: opening log /var/log/squid/access.log
> 2013/02/15 8:30:10| Unlinkd pipe opened on FD 12
> 2013/02/15 8:30:10| Swap maxSize 8192 + 8388608 KB, estimated 645907 objects
> 2013/02/15 8:30:10| Target number of buckets: 32295
> 2013/02/15 8:30:10| Using 32768 Store buckets
> 2013/02/15 8:30:10| Max Mem size: 8388608 KB
> 2013/02/15 8:30:10| Max Swap size: 8192 KB
> 2013/02/15 8:30:10| Local cache digest enabled; rebuild/rewrite every
> 3600/3600 sec
> 2013/02/15 8:30:10| logfileOpen: opening log /var/log/squid/store.log
> 2013/02/15 8:30:10| Rebuilding storage in /var/spool/squid (CLEAN)
> 2013/02/15 8:30:10| Using Least Load store dir selection
> 2013/02/15 8:30:10| Set Current Directory to /var/spool/squid
> 2013/02/15 8:30:10| Loaded Icons.
> 2013/02/15 8:30:10| Accepting accelerated HTTP connections at 0.0.0.0,
> port 80, FD 14.
> 2013/02/15 8:30:10| Accepting ICP messages at 0.0.0.0, port 3130, FD 15.
> 2013/02/15 8:30:10| HTCP Disabled.
> 2013/02/15 8:30:10| WCCP Disabled.
> 2013/02/15 8:30:10| Configuring localhost Parent localhost/12080/0
> 2013/02/15 8:30:10| Ready to serve requests.
> 2013/02/15 8:30:10| Done reading /var/spool/squid swaplog (11 entries)
> 2013/02/15 8:30:10| Finished rebuilding storage from disk.
> 2013/02/15 8:30:10| 11 Entries scanned
> 2013/02/15 8:30:10| 0 Invalid entries.
> 2013/02/15 8:30:10| 0 With invalid flags.
> 2013/02/15 8:30:10| 11 Objects loaded.
> 2013/02/15 8:30:10| 0 Objects expired.
> 2013/02/15 8:30:10| 0 Objects cancelled.
> 2013/02/15 8:30:10| 0 Duplicate URLs purged.
> 2013/02/15 8:30:10| 0 Swapfile clashes avoided.
> 2013/02/15 8:30:10| Took 0.3 seconds ( 41.8 objects/sec).
> 2013/02/15 8:30:10| Beginning Validation Procedure
> 2013/02/15 8:30:10| Completed Validation Procedure
> 2013/02/15 8:30:10| Validated 11 Entries
> 2013/02/15 8:30:10| store_swap_size = 44k
> 2013/02/15 8:30:11| storeLateRelease: released 0 objects
> 2013/02/15 8:30:35| CACHEMGR: <unknown>@127.0.0.1 requesting 'info'
> 2013/02/15 8:30:39| CACHEMGR: <unknown>@127.0.0.1 requesting 'info'
> 2013/02/15 8:30:40| CACHEMGR: <unknown>@127.0.0.1 requesting 'info'
> 2013/02/15 8:30:42| CACHEMGR: <unknown>@127.0.0.1 requesting 'info'
> 2013/02/15 8:30:44| CACHEMGR: <unknown>@127.0.0.1 requesting 'info'
> 2013/02/15 8:33:10| CACHEMGR: <unknown>@127.0.0.1 requesting 'info'
>
> when connection timeout error occurred, I do not see error about file
> descriptor.
>
> my /etc/sysclt.conf
> net.ipv4.tcp_syncookies = 1
> net.ipv4.tcp_tw_reuse = 1
> net.ipv4.tcp_tw_recycle = 1
> net.ipv4.tcp_fin_timeout = 30
> fs.file-max = 65536
>
> my squid.conf has:
> max_filedescriptors 2048
>
> and my squidclient says:
> squidclient -p 80 mgr:info | grep "file desc"
> Maximum number of file descriptors: 2048
> Available number of file descriptors: 1651
> Reserved number of file descriptors: 100

There you go then. Squid is not permitted to _use_ more than 1651 FD.
Every client TCP connection uses at least 1, sometimes 2 FD.
When all the FD are used up Squid waits until some are free'd before
accepting more client connections.

With "from 4651 to 9404" I would set your max_filedescriptors to at
least 18000. It can be anything up to the ulimit max.

Amos
Received on Fri Feb 15 2013 - 12:29:37 MST