[2515] in SIPB-AFS-requests
Re: Potential SIPB cell outage Friday 10/25/96 at 1700
daemon@ATHENA.MIT.EDU (Sam Hartman)
Fri Oct 25 15:37:28 1996
To: mhpower@MIT.EDU
Cc: sipb-afsreq@MIT.EDU
From: Sam Hartman <hartmans@MIT.EDU>
Date: 25 Oct 1996 15:35:33 -0400
In-Reply-To: mhpower@MIT.EDU's message of Fri, 25 Oct 1996 01:38:03 EDT
>>>>> "mhpower" == mhpower <mhpower@MIT.EDU> writes:
mhpower> from [2502]:
>> Assuming the new kernel continues to work, I think this patch
>> should be installed on all of the other 4.1.3_U1
>> machines. Presumably this should happen on reynelda before
>> reynelda becomes a database server, to minimize the likely
>> impact of the outage.
mhpower> So, will the outage on Friday allow for a reboot of
mhpower> reynelda with a new kernel,
This could be done but I think it would be a bad idea because
it would significantly increase the impact of the outage; we would
actually have a fileserver down for a few minutes. I don't think it
would be appropriate to do this in mid afternoon.
mhpower> do people think that it's a
mhpower> better idea to reboot as part of a much more disruptive
mhpower> later outage that takes down a db server, or is the
mhpower> feeling that the patch is unnecessary (maybe it's
mhpower> expected that getsockopt will never be called on
mhpower> reynelda)?
Taking down one of three db servers is not much more
disruptive than taking down one fileserver. Quorum will be
maintained. You will get an election when you take down renelda, but
that's not too major compared to taking down a fileserver in a cell
with only three fileservers.
Also, I don't think renelda should call getsockopt very often
at all.
mhpower> Matt
