[OpenAFS] Re: Re: afs db failover testing for lowest IP

[Stephen Joyce]

Hey, John.

Let me know if you have more problems, or just need to bounce any ideas 
around.

I went through something similar last summer, but I actually changed IP 
addresses (moved to a different VLAN). I decided to bite the bullet and 
virtualize my DB servers at the same time. You may want to consider the 
same. It has the potential to make some future tasks much simpler (esp. 
disaster recovery).

I think I ended up having 3 short downtimes (announced 1 hour; finished in 
15 mins each time) scheduled during semester breaks.

For safety, I've currently got all 3 DB servers are on separate dedicated, 
low-power hypervisors in the same cluster (Dell R410s), so I'm not really 
eliminating physical boxes per-se, but it gives me flexibility to snap the 
VMs, reboot *fast*, live migrate the VMs between HVs for patching, hardware 
upgrades, etc...

Stephen

On Wed, 24 Jan 2018, John Sopko wrote:

> I did not have the secondary db machines in /usr/vice/etc/CellServDB.
> All is good :)
>
> On Wed, Jan 24, 2018 at 9:57 AM, John Sopko <sopko@cs.unc.edu> wrote:
>> I need to upgrade our afs db servers. It has been sometime so I setup
>> a test cell with 3 servers to see how things react when the lowest IP
>> server is down.
>>
>> When the lowest IP db server is down udebug shows the second lowest IP
>> becomes the sync site. But, I cannot do certain commands that time out
>> because they are trying to use the lowest IP address that id down. For
>> example I cannot do "pts listentries -users" "vos listvldb". The file
>> services seem to be ok.
>>
>> I use "strace pts listentires -users" which shows the client keeps
>> trying the down machines IP until it finally times out. Can anyone
>> shed any light on what is going on. Thanks.
>>
>>
>> --
>> John W. Sopko Jr.
>> University of North Carolina
>> Computer Science Dept CB 3175
>> Chapel Hill, NC 27599-3175
>>
>> Fred Brooks Building; Room 140
>> Computer Services Systems Specialist
>> email: sopko AT cs.unc.edu
>> phone: 919-590-6144
>
>
>
> -- 
> John W. Sopko Jr.
> University of North Carolina
> Computer Science Dept CB 3175
> Chapel Hill, NC 27599-3175
>
> Fred Brooks Building; Room 140
> Computer Services Systems Specialist
> email: sopko AT cs.unc.edu
> phone: 919-590-6144
> _______________________________________________
> OpenAFS-info mailing list
> OpenAFS-info@openafs.org
> https://lists.openafs.org/mailman/listinfo/openafs-info
>

--
Stephen Joyce
Linux Systems Specialist
Office of Arts & Sciences Information Services
University of North Carolina at Chapel Hill