Users of RBL's
Ken A
ka at pacific.net
Tue Jun 27 16:56:02 IST 2006
Chris Hammond wrote:
>>>> Ken A <ka at pacific.net> 06/26/06 6:29 PM >>>
>
>
>>> Chris Hammond wrote:
>>> I am running a caching bind server. I found rbldnsd but everything I see points to using it
>>> with local hard copies of the rbl itself that has to be rsync'd from the rbl provider. I don't
>>> understand it enough to be able to figure out if it can operate like bind in caching mode.
>>> I am looking at different areas to try and determine what where my bottleneck is.
>>>
>>> It does not appear to be memory, the machine has 1.5GB of that.
>
>> what does 'free' say about swap in use. 1.5gb may not be enough,
>> depending on how many child processes of MailScanner you are running,
>> and how much ram everything else you have going uses.
>
> Just for giggles, I turned off the swap partitions 2 days ago and here is what free reports.
>
> total used free shared buffers cached
> Mem: 3116372 2908432 207940 0 398888 1513984
> -/+ buffers/cache: 995560 2120812
> Swap: 0 0 0
>
> I am using the standard 5 child processes recommended per processor and have only one processor.
>
> The processor is an Opteron
>> 242 (1.6Ghz) and it doesn't seem to be the issue. The system is running a caching bind server.
>> I also have razor2, pyzor, rules_du_jour (none of the BIG nasty ones). I am having 30 message
>> batch times of 180- 280 seconds. This is a single server running everything including Mailwatch
>> and mysql database which I have used mysqlard to try and tune. I turned on MCP over the weekend
>> and my batch times jumped to 680+ seconds. Obviously that wasn't going to work. But now, I
>> am looking at another possibility. Drive subsystem. The server is an HP Proliant DL145 with a
>> pair of 80GB IDE drives software mirrored.
>
>> On different controllers, or the same?
>
> Different controllers, hda and hdc
>
>> What does 'vmstat 2' say?
>
> procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
> r b swpd free buff cache si so bi bo in cs us sy id wa
> 0 0 0 296436 399156 1482420 0 0 36 102 84 96 69 11 20 1
> 1 0 0 256620 399160 1482440 0 0 0 64 1086 190 23 8 70 0
> 1 0 0 207476 399180 1495972 0 0 4 222 1194 309 55 11 31 4
> 2 0 0 196692 399188 1482736 0 0 0 5444 1144 230 90 10 0 0
> 3 0 0 179620 399188 1482740 0 0 0 70 1115 328 90 10 0 0
> 3 0 0 249796 399196 1493560 0 0 0 0 1114 232 87 13 0 0
> 2 1 0 224188 399196 1493340 0 0 0 5386 1124 184 92 8 0 0
> 4 0 0 241092 399204 1483324 0 0 30 324 1477 790 79 21 0 1
> 3 0 0 189948 399208 1494032 0 0 0 344 1134 211 84 17 0 0
>
>> How about 'iostat - x' ?
>
> Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s rkB/s wkB/s avgrq-sz avgqu-sz await svctm %util
> hda 0.38 339.96 1.23 13.97 35.02 2836.38 17.51 1418.19 189.00 1.13 74.27 3.70 5.62
> hdc 0.70 339.97 1.22 13.96 36.53 2836.38 18.26 1418.19 189.19 1.15 75.33 3.73 5.67
> md0 0.00 0.00 0.92 0.00 1.84 0.00 0.92 0.00 2.00 0.00 0.00 0.00 0.00
> md1 0.00 0.00 2.61 353.35 69.69 2826.78 34.84 1413.39 8.14 0.00 0.00 0.00 0.00
>
>> If you don't have it, 'yum install sysstat'
>
> Already there.
>
>> Sounds like you may just be asking too much of the hardware.
>
> This could very well be. Before I go asking for a new server though, I want to make sure I have my ducks in a row.
> When this was nothing more than a Postfix box with static rules, it handled the job just fine. But I think it may
> be really working for it's living.
MailScanner and SpamAssassin do use a lot of resources. It looks to be
cpu bound, but that's a good thing usually! Any way to upgrade that
processor? To reduce CPU usage, tune/configure some software. Did you
read the performance tweaks section in the mailscanner wiki? To reduce
disk writes, setup syslog to log to another box, or put mysql on another
box, or throw another cheap ide drive into the box and log to it,
instead of the mirrored drives.
Ken A.
Pacific.Net
>
> Thanks
> Chris
>
More information about the MailScanner
mailing list