Load balancers usually have fairly minimal scripted protocol tests, just enough to confirm probable correct operation, but frequently omit protocol niceties, or don't perform any useful operation, which is what sendmail is complaining about. The SMTP default is to just send
QUIT (checking the response codes for each).
This is a LogLevel > 5 message, the default is LogLevel 9, so you could either recreate your
.cf after adding to the
or temporarily override it by adding
-O LogLevel=5 on the sendmail command line. The downside being many other (~140) messages over levels 6-9 will also be suppressed.
There is only one documented scenario where that specific message is suppressed — when a connection is going to be rejected anyway via configuration (access map or TCP wrappers). However every scenario I can think of involves logging a rejection error instead... (and worse, some may return an error to the
HELO greeting which would cause the F5 to mark the service down).
There is no configurable item in the standard
SMTP_monitor, but as a workaround you could set up an alternate SMTP monitor script, this example on F5's devcentral (registration required to view source) shows how it's done using Expect:
That example sends an email on each probe which you almost certainly don't want ;-)
Instead try something like this:
expect "220" ;# must wait for SMTP banner
send "HELO mydomain.com\r\n" ;# EHLO is better
send "VRFY postmaster\r\n" ;# some non-NOP command
send "QUIT\r\n" ;# clean exit
and adjust as required. If your sendmail is fussy (
PrivacyOptions might block
ETRN) then either amend the expected code, or use
send "MAIL from:<email@example.com>\r\n"