On 8/15/12 2:24 PM, Brandon Allbery wrote:
Those *are* the DB servers (minus kaserver, but if you're using aklog
you don't care and if you *are* still running kaserver for some reason
then you should find a way to get rid of it ASAP). AFSDB was
deprecated by the OpenAFS community when those SRV records became
standard; if anything, you should be replacing it with the SRV records.
I would suggest using strace (or similar OS-specific utility such as
dtruss on OS X and Solaris or ktrace on *BSD) to find out exactly
where aklog is hanging up. If it's hanging up looking for a
nonexistent AFSDB RR, then it's actually a problem with your DNS
server not returning the appropriate response to a failed query, and
the workaround if you cannot get the DNS servers to actually speak the
DNS protocol properly might be to rebuild OpenAFS without AFSDB
support, or possibly run local caching nameservers. (Although I would
think it'd check for the correct SRV RRs first and AFSDB only as a
fallback.)
Also, make sure the SRV records are what they're supposed to be.
_afs3-prserver._udp.$cellname for whatever the cellname is. (Note,
not the domain, not the Kerberos realm, *certainly* not a hostname.)
Okay, I fixed the SRV records (I had one entry for each server instead
of just the DB servers alone) and now things seem to be moving faster.
Thanks for the info.
_______________________________________________
OpenAFS-info mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-info
