[OpenAFS] Re: cyrus with storage in afs?

Derrick Brashear shadow@gmail.com
Tue, 26 Jun 2007 15:25:19 -0400


------=_Part_28791_31200115.1182885919585
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

On 6/26/07, Adam Megacz <megacz@cs.berkeley.edu> wrote:
>
>
> Derrick J Brashear <shadow@dementia.org> writes:
> >> For argument's sake, let's assume only a single front-end (imapd)
> >> machine.
>
> > That's not especially dangerous, then. the worst risk is screwing with
> > Cyrus' idea of the universe if a volume goes away under it, like, in
> > particular i'm uncertain we ever audited cyrus to see if it handled
> > write failures correctly in every case,
>
> Ok, but this should be no worse than (a misguided effort at) running
> cyrus out of NFS, right?
>
> > I suppose I should say the locking semantics we have do make having
> > multiple front ends dangerous.
>
> Sure, I can see that.
>
> So, is it safe to say that running cyrus out of AFS is no more risky
> than running cyrus out of NFS?  I think that would be a whole lot less
> mysterious.


It's no more risky than that, at least for the one frontend case. Otherwise,
I think it's a case of different risks.

------=_Part_28791_31200115.1182885919585
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

<br><br><div><span class="gmail_quote">On 6/26/07, <b class="gmail_sendername">Adam Megacz</b> &lt;<a href="mailto:megacz@cs.berkeley.edu">megacz@cs.berkeley.edu</a>&gt; wrote:</span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<br>Derrick J Brashear &lt;<a href="mailto:shadow@dementia.org">shadow@dementia.org</a>&gt; writes:<br>&gt;&gt; For argument&#39;s sake, let&#39;s assume only a single front-end (imapd)<br>&gt;&gt; machine.<br><br>&gt; That&#39;s not especially dangerous, then. the worst risk is screwing with
<br>&gt; Cyrus&#39; idea of the universe if a volume goes away under it, like, in<br>&gt; particular i&#39;m uncertain we ever audited cyrus to see if it handled<br>&gt; write failures correctly in every case,<br><br>Ok, but this should be no worse than (a misguided effort at) running
<br>cyrus out of NFS, right?<br><br>&gt; I suppose I should say the locking semantics we have do make having<br>&gt; multiple front ends dangerous.<br><br>Sure, I can see that.<br><br>So, is it safe to say that running cyrus out of AFS is no more risky
<br>than running cyrus out of NFS?&nbsp;&nbsp;I think that would be a whole lot less<br>mysterious.</blockquote><div><br>It&#39;s no more risky than that, at least for the one frontend case. Otherwise, I think it&#39;s a case of different risks.
<br>&nbsp;<br></div><br></div><br>

------=_Part_28791_31200115.1182885919585--