[OpenAFS] Re: vldb_check -servers cleanup and empty server entry messages

John W. Sopko Jr. sopko@cs.unc.edu
Fri, 05 Mar 2010 08:41:26 -0500


Thanks for the info. I am going to delete the
rest of the non MH machines. One more question.
Should the vldb database contain only file server
machines? One of the MH entries is our Kerberos 5
machine, it is not a AFS file or db server.
Does that need to be in the vldb db? The
k5 entry may have gotten in there when I was doing
testing/moving to K5 years ago.

Thanks for your support.

Andrew Deason wrote, On 3/4/2010 12:24 PM:
> On Thu, 25 Feb 2010 16:08:46 -0500
> "John W. Sopko Jr."<sopko@cs.unc.edu>  wrote:
>
>> % vldb_check /usr/afs/db/vldb.DB0 -servers |&  head -40
>> VLDB_CHECK_WARNING: Ubik header size is 0 (should be 64)
>> MH block 0, index 1: 152.2.128.4
>> MH block 0, index 3: 152.2.128.3
>> MH block 0, index 4: 152.2.129.145
>> MH block 0, index 31: 152.2.129.25
>> MH block 0, index 32: 152.2.128.34
>>      Server ip addr 4 = 152.2.128.157
>>      Server ip addr 5 = 152.2.128.161
> [...]
>> The entries like "Server ip addr ..." our not file or db servers and in
>> some cases are not in DNS. These IPs are owned by us but our not
>> file servers.
>
> The entries like 'Server ip addr ...' are non-multihomed entries. All
> modern OpenAFS fileservers will register themselves as multihomed, so
> non-multihomed entries will typically be a result of either a really old
> fileserver, or someone running 'vos changeaddr' without -remove.
>
> Really old fileservers aren't really anything to worry about; they just
> never go away in the database unless you delete them (as you did). It
> could also be someone doing a 'vos changeaddr', though; make sure you
> don't use that command without the -remove flag (as documented, using it
> except to remove addresses isn't a good idea).
>
>> Previous to this vldb_check I ran another check and there were 4
>> "server ip addr ..." entries that were not in the IP range we own or
>> DNS.  I used "vos changeadress 128.109.136.161 -remove" to remove the
>> 4 entries.
>>
>> I then ran vldb_check /usr/afs/db/vldb.DB0 -servers again, the entries
>> were removed but I now get a bunch of "empty server entry 0" entries!
>> A sample of a few entries is shown below. The volumes that are
>> displayed seem to be fine. It is making me nervous!
>
> This is nothing to worry about; this is a bug in vldb_check. vldb_check
> doesn't read in the entries properly, so it thinks server numbers are
> always 0. If you have a server in index 0, it doesn't think there's a
> problem.  If you don't have a server in index 0, it thinks the entry is
> pointing to something nonexistent, which is a problem. You happened to
> delete the server entry in slot 0, so now it's whining at you. But
> there's nothing wrong with doing that; vldb_check just isn't reading the
> entries right.
>
> I just filed ticket 126661 about this; follow it if you want the fix.
> But for now, it's nothing to worry about.
>

-- 
John W. Sopko Jr.               University of North Carolina
email: sopko AT cs.unc.edu      Computer Science Dept., CB 3175
Phone: 919-962-1844             Fred Brooks Building; Room 140
Fax:   919-962-1799             Chapel Hill, NC 27599-3175