Re: if /var/spool greater than /var

I have two hosts, one is an incoming MTA/POP3 server with IP address
192.168.0.252, other is an outgoing MTA server with IP address 192.168.0.247.
they are running the same postfix 2.2.3.

My problem is, there are 90% outgoing messages are delivered to inbound side.
but recently I have been seeing the following in my logs in the outbound site:

Jun 2 11:18:30 fb postfix/qmgr[80084]: 7AEAB67883:
to=<patrickchu@xxx.xxx>, relay=none, delay=5734,
status=deferred (delivery temporarily suspended: connect to 192.168.0.252[192.168.0.252]: Connection refused)

when the deferred queue is growing large, I done a "postqueue -f"
on outbound side (bad behavior), this often cause the inbound side
sudden running slowly, the disk red-light is flashing, if it still doesn't turn green, I end up
doing a 'postfix stop/restart' to slove this problem.

The following is the output of Postfix log summaries:

Grand Totals
------------
messages

7552 received
25192 delivered
1851 forwarded
19 deferred (22 deferrals)
380 bounced
6561 rejected (20%)
9272 reject warnings
563 held
0 discarded (0%)

516344k bytes received
3638m bytes delivered
1367 senders
780 sending hosts/domains
930 recipients
285 recipient hosts/domains

smtpd

9456 connections
730 hosts/domains
14 avg. connect time (seconds)
35:39:48 total connect time


Per-Hour Traffic Summary
time received delivered deferred bounced rejected
--------------------------------------------------------------------
0000-0100 183 215 0 11 521
0100-0200 189 96 0 17 400
0200-0300 177 74 1 25 380
0300-0400 120 68 0 8 329
0400-0500 141 121 0 14 320
0500-0600 137 131 2 13 459
0600-0700 161 140 0 11 361
0700-0800 148 118 0 15 359
0800-0900 236 424 0 7 705
0900-1000 466 2082 4 35 909
1000-1100 495 2308 1 16 997
1100-1200 488 2102 0 9 1049
1200-1300 338 1570 1 46 705
1300-1400 321 1416 0 14 706
1400-1500 463 2164 2 10 996
1500-1600 447 1853 3 15 817
1600-1700 668 3035 2 5 1271
1700-1800 549 2440 0 20 1068
1800-1900 675 2137 2 21 1297
1900-2000 277 867 1 14 578
2000-2100 227 572 0 11 555
2100-2200 300 660 0 20 740
2200-2300 172 279 3 10 486
2300-2400 174 320 0 13 388


in 'man qmgr', it shows:

A single queue manager process has to compete for disk access with
multiple front-end processes such as cleanup(8). A sudden burst of
inbound mail can negatively impact outbound delivery rates.


and I also readed the bottleneck analysis page, maybe I misconfig something,
the inbound mail still doesn't work well. (the outbound mail works well greatly.)


Victor Duchovni wrote:

>On Wed, Jun 01, 2005 at 09:05:32AM +0800, yeskw@ms15.hinet.net wrote:
>
>>Thank you, Here is the output of iostat -x 10, what's wrong with it?
>>
>># iostat -x 10
>> extended device statistics
>>device mgr/s mgw/s r/s w/s kr/s kw/s size queue wait
svc_t
>>%b
>>hda 1 20 1.7 6.6 41.6 107.0 17.9 1.4 173.4
1.7
>>1
>>...
>
>
>This does not look like an I/O bound system. Whatever performance
problems
>you may be having are elsewhere. Are you in fact experiencing performance
>issues? What are they?
>