PfSense 2.2.3 - Internet is very slow via Squid3

heper · Jun 29, 2015, 4:10 PM

fterwards i installed Squid3, SquidGuard, LightSquid, Sarg, Snort, VHosts, Cron,bandwidthD. ntopng

none of the above are known for their stability and reliability (especially, short after a release).

I would start with one package at a time and see what the culprit is.

most likely culprits: LightSquid, Sarg, SquidGuard, VHosts, bandwidthD

KOM · Jun 29, 2015, 4:38 PM

Focus on squid3 and snort. I can't possibly imagine how a log analyzer like Lightsquid or Sarg would slow down his web browsing since they only run ever now and then.

gdsnytech · Jun 29, 2015, 8:20 PM

I have removed all of the packages listed and tried installing squid 2 but i'm still experiencing the same lag.

Is it possible that it could be a disk read error issue?

jimp · Jun 29, 2015, 8:25 PM

Some have found that the disk changes we made for sync mode have slowed down squid by default.

If you want to risk the possibility of disk corruption but gain speed, edit the ,sync out of /etc/fstab for the root slice and/or run

mount -o nosync /

KOM · Jun 29, 2015, 8:55 PM

Squid2 is crap and very old. Stick to squid3.

gdsnytech · Jun 30, 2015, 1:30 AM

@jimp:

Some have found that the disk changes we made for sync mode have slowed down squid by default.

If you want to risk the possibility of disk corruption but gain speed, edit the ,sync out of /etc/fstab for the root slice and/or run
mount -o nosync /

I forgot to mention that the pfsense install is on a RAID 1

but that command seems to have did it. HTTP requests seems to be more responsive! THANK YOU SIR! ;-)

gdsnytech · Jun 30, 2015, 6:35 PM

How do i make this mount option stick on fstab? After reboot the mount -o nosync option goes away.

doktornotor · Jun 30, 2015, 8:12 PM

@gdsnytech:

How do i make this mount option stick on fstab? After reboot the mount -o nosync option goes away.

Would help to re-read the post quoted. It's already written there.

gdsnytech · Jul 1, 2015, 12:05 AM

That is what i did. For some reason it wasn't sticking after rebooting. I had to keep going into /etc/fstab to make the change. But it seems to be sticking now.

hbc · Jul 1, 2015, 7:53 AM

@jimp:

Some have found that the disk changes we made for sync mode have slowed down squid by default.

If you want to risk the possibility of disk corruption but gain speed, edit the ,sync out of /etc/fstab for the root slice and/or run
mount -o nosync /

Could this change to disk sync also be the issue for my listening queue problem with squid? –> https://forum.pfsense.org/index.php?topic=95873.0

At least as soon as I disable disk sync, the listen queue size drops to zero. I compared my /etc/fstab from backup with current one and this sync options wasn't present before.

Maybe this option creates big performance impact on RAID systems. My secondary CARP node - also running 2.2.3 and sync option enabled - does not have this problem, but doesn't have RAID either.

gdsnytech · Jul 1, 2015, 6:39 PM

@hbc:

Could this change to disk sync also be the issue for my listening queue problem with squid? –> https://forum.pfsense.org/index.php?topic=95873.0

At least as soon as I disable disk sync, the listen queue size drops to zero. I compared my /etc/fstab from backup with current one and this sync options wasn't present before.

Maybe this option creates big performance impact on RAID systems. My secondary CARP node - also running 2.2.3 and sync option enabled - does not have this problem, but doesn't have RAID either.

It is definitely a RAID issue. Squid caching performance suck on a RAID. Especially with the 'mount sync' option. Just edit your /etc/fstab file with mount option 'nosync' with 'Edit File' in 'Diagnostics'. Save and reboot the box. When it comes back check fstab to make sure that the change is still there.

Other mod changes that i made were to 'Systsm> Advance> System Tunables>'

Tunable name Value
vfs.read_max from 32 to 128

https://doc.pfsense.org/index.php/Squid_Package_Tuning

Created
kern.ipc.nmbclusters 32768

created the above ONLY if you are using the 'diskd' 'Hard disk cache system' setup in Squid. Which requires you to copy ipcs and ipcrm from a FreeBSD 10.1 ISO /usr/bin/ to pfsense /usr/local/bin

also reboot and check after reboot to make sure that the settings are still there.

Squid is 'flying right now'. Well sort of. But the 'nosync' boot option in fstab does make a BIG different.

I am running squid in a production environment.

mesro09 · Jul 4, 2015, 6:25 PM

i have already this problem but i dont have any raid config
everything was good working until i update from 2.2.2 to 2.2.3 and intennet put so slow now i am finding solution i already tried to reinstall squid and squidguard but nothing is solved now i dont have any filter
i installed new one with fresh install pfsense 2.2.3 but with fresh install i have problem too after restart pfsense squid and squidguard stops and cant start again (squid-1): The redirector helpers are crashing too rapidly, need help!
my system was squid3 and squidguard-dev amd64 squid non transparent (with wpad)
i hope to solve
second time i did big problems (first when i update to 2.2.2 rebooting time delay so much) and 2.2.3 squid problems
i think it is time to change FW with others

gdsnytech · Jul 7, 2015, 9:28 AM

@mesro09:

i have already this problem but i dont have any raid config
everything was good working until i update from 2.2.2 to 2.2.3 and intennet put so slow now i am finding solution i already tried to reinstall squid and squidguard but nothing is solved now i dont have any filter
i installed new one with fresh install pfsense 2.2.3 but with fresh install i have problem too after restart pfsense squid and squidguard stops and cant start again (squid-1): The redirector helpers are crashing too rapidly, need help!
my system was squid3 and squidguard-dev amd64 squid non transparent (with wpad)
i hope to solve
second time i did big problems (first when i update to 2.2.2 rebooting time delay so much) and 2.2.3 squid problems
i think it is time to change FW with others

Just try the suggestion above and make the change on fstab.

Mr. Jingles · Jul 7, 2015, 9:35 AM

@jimp:

Some have found that the disk changes we made for sync mode have slowed down squid by default.

If you want to risk the possibility of disk corruption but gain speed, edit the ,sync out of /etc/fstab for the root slice and/or run
mount -o nosync /

My squid3 is not fast either. But I don't like risking disk corruption, possibly destroying pfSense. That is: what are the odds disk corruption will appear, Jim? Relevant variables? A thumb number (0,005% or 60%)?

KOM · Jul 7, 2015, 1:23 PM

DNS issues can also make squid look slow. Shell in and run:

squidclient -h LAN_IP_Address -p 3128 mgr:info

Then read the report, paying special attention to the Median Service Times section. Look for anything that seems large as compared to the others.

Mr. Jingles · Jul 7, 2015, 3:26 PM

@KOM:

DNS issues can also make squid look slow. Shell in and run:

squidclient -h LAN_IP_Address -p 3128 mgr:info

Then read the report, paying special attention to the Median Service Times section. Look for anything that seems large as compared to the others.

I do not want to hijack this thread, so if I have to create a new thread (similar problem), I will, just let me know :-[

[quote]
squidclient -h localhost -p 3128 mgr:info
Sending HTTP request … done.
HTTP/1.1 200 OK
Server: squid
Mime-Version: 1.0
Date: Tue, 07 Jul 2015 15:23:33 GMT
Content-Type: text/plain
Expires: Tue, 07 Jul 2015 15:23:33 GMT
Last-Modified: Tue, 07 Jul 2015 15:23:33 GMT
X-Cache: MISS from squid
X-Cache-Lookup: MISS from squid:3128
Connection: close

Squid Object Cache: Version 3.4.10
Build Info:
Start Time: Tue, 07 Jul 2015 13:39:21 GMT
Current Time: Tue, 07 Jul 2015 15:23:33 GMT
Connection information for squid:
Number of clients accessing cache: 2
Number of HTTP requests received: 1037
Number of ICP messages received: 0
Number of ICP messages sent: 0
Number of queued ICP replies: 0
Number of HTCP messages received: 0
Number of HTCP messages sent: 0
Request failure ratio: 0.00
Average HTTP requests per minute since start: 10.0
Average ICP messages per minute since start: 0.0
Select loop called: 609439 times, 10.259 ms avg
Cache information for squid:
Hits as % of all requests: 5min: 23.7%, 60min: 12.2%
Hits as % of bytes sent: 5min: 37.0%, 60min: 4.2%
Memory hits as % of hit requests: 5min: 0.0%, 60min: 8.9%
Disk hits as % of hit requests: 5min: 0.0%, 60min: 8.9%
Storage Swap size: 6668 KB
Storage Swap capacity: 0.0% used, 100.0% free
Storage Mem size: 4620 KB
Storage Mem capacity: 0.2% used, 99.8% free
Mean Object Size: 12.97 KB
Requests given to unlinkd: 0
Median Service Times (seconds) 5 min 60 min:
HTTP Requests (All): 0.10857 0.32154
Cache Misses: 0.12783 0.37825
Cache Hits: 0.00000 0.07014
Near Hits: 0.00000 0.22004
Not-Modified Replies: 0.05633 0.05633
DNS Lookups: 0.01940 0.02683
ICP Queries: 0.00000 0.00000
Resource usage for squid:
UP Time: 6252.187 seconds
CPU Time: 73.719 seconds
CPU Usage: 1.18%
CPU Usage, 5 minute avg: 0.89%
CPU Usage, 60 minute avg: 1.27%
Maximum Resident Size: 178544 KB
Page faults with physical i/o: 0
Memory accounted for:
Total accounted: 7844 KB
memPoolAlloc calls: 246629
memPoolFree calls: 255812
File descriptor usage for squid:
Maximum number of file descriptors: 58977
Largest file desc currently in use: 51
Number of file desc currently in use: 32
Files queued for open: 0
Available number of file descriptors: 58945
Reserved number of file descriptors: 100
Store Disk files open: 0
Internal Data Structures:
569 StoreEntries
537 StoreEntries with MemObjects
536 Hot Object Cache Items
514 on-disk objects

What would you make of this, KOM?

Thank you :P

KOM · Jul 7, 2015, 4:19 PM

Everything looks normal. I think your problem is disk-related as you suspected.

gdsnytech · Jul 7, 2015, 6:35 PM

@Mr.:

@jimp:
Some have found that the disk changes we made for sync mode have slowed down squid by default.

If you want to risk the possibility of disk corruption but gain speed, edit the ,sync out of /etc/fstab for the root slice and/or run
mount -o nosync /
My squid3 is not fast either. But I don't like risking disk corruption, possibly destroying pfSense. That is: what are the odds disk corruption will appear, Jim? Relevant variables? A thumb number (0,005% or 60%)?

I have made the change in fstab both for work (two sites) and home setup and it is working fine. Just make the change in fstab and then reboot.

mesro09 · Jul 10, 2015, 8:19 PM

hello
let me explain you this changes for squid caches but i dont use squid for cache contens my config is 0
so can i make this changes ?¿?

chidgear · Aug 28, 2015, 8:44 PM

Hi!. I'm having this same issue, when the proxy is enabled the webpages takes a lot of time to load and, once loaded it becomes more responsive but, still, is very slow.
I have installed Squid3+SquidGuard Transparent proxy enabled and SSL filtering on. I've disabled squidguard (just to make sure) and the issue keeps there. Then, when I configured the lannet to bypass the proxy, the issue dissapeared so I think I've missed something important with Squid3.
I tried to edit the /etc/fstab from this:

# Device		Mountpoint	FStype	Options		Dump	Pass#
/dev/ufsid/558c431cbd7f951e		/		ufs	rw,sync		1	1
/dev/label/swap0		none		swap	sw		0	0

removing the ,sync so the file now reads:

# Device		Mountpoint	FStype	Options		Dump	Pass#
/dev/ufsid/558c431cbd7f951e		/		ufs	rw		1	1
/dev/label/swap0		none		swap	sw		0	0

I ran the command

mount -o nosync /

and restarted the box…
The issue? still there...
Am I doing something wrong?

Thanks in advance

-------Update------------

I got tired so did a full restauration from a file that I've backed up previously. I didn't knew what was causing my trouble, but now it is gone. I suggest the young adventurers that they do, at leas one time at week, a full backup of their configurations so, if they experience some isssue of this kind, have where to "run"... It is easier and faster.