I have one SF (called SF1) in DMZ doing only IP/DNS tests. It forwards to another SF (called SF2) that runs other content tests as needed. SF1 has a white list and uses Auth-TO list. SF2 does not have a whitelist and uses identical Auth-To list.
99% of time this config works without issues.
Once in a while there is a huge lock up (100% CPU usage) caused by very specific bug in SF program? or sender’s email heather perhaps? or ?
In same specific Unknown scenario, SF1 receives email correctly and sends it to SF2, but SF2 reports that “no relaying is permitted and % found in from address” and rejects email, both unfound.
SF2 waits until read-timeout has reached to do this Error reporting. When SF2 does so, it shows "from" address field is empty and reports what was originally in “From” field (in SF1) as “To” address instead!!! This perhaps explains relay error as “To” user in this case is no longer “From” rcpt to user which obviously also is not in expected Auth_to list. And message should be rejected.
My question is how can I investigate this error?
In each case SF1 reports pending message in outbound queue.
When read time-out of SF2 is reached, SF1 reports “message was returned to sender - server error - Read Timeout” as well as “- forwarding to xxxx: - server error - Read Timeout
SF2 reports message stuck in inbound queue at “PROCESSING DATA…” state
When read timeout of SF2 is reached, SF2 reports email is rejected - no relay allowed or % found
As well as a - Notice - IdleDisconnectMinutesTimeout reached. Removing threadID.
Read timeout is rage such as 15minutes. And server is not heavily loaded. Plenty of ram and very fast 2.4GHZ CPU.
Obviously on big is when SF1 sends to SF2, from address in SF1 becomes “TO” address in SF2 and “from” address is reported as empty. SF2 fails to grab “From” address and eventually times out.
Another observation is that in all such scenario CPU usages pegs to 100%. I thing SF trying to parse something…
This scenario occurs once in a while... and normally occurs with inbound messages from
Mailing lists such as CNet ones. There are many others as well. Do these have special header content? Or is spamfilter problem?
In all cases, I noticed, SF2 has temp files generated in temp directory of SF2
And all seams to have tmp extension as expected. I opened a few, but all files are GIF98a attachments (could be co incidence). However, queue directory of SF2 remains empty.
Any assistance is appreciated to debug this.
|