[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Pan-users] Proper behavior of "backup" servers?
From: |
Duncan |
Subject: |
Re: [Pan-users] Proper behavior of "backup" servers? |
Date: |
Sun, 20 Nov 2011 21:12:20 +0000 (UTC) |
User-agent: |
Pan/0.135 (Tomorrow I'll Wake Up and Scald Myself with Tea; GIT 6ffb80b /st/portage/src/egit-src/pan2) |
walt posted on Sun, 20 Nov 2011 16:29:41 +0000 as excerpted:
> While testing pan's new ssl function I discovered that setting gmane.org
> to "backup server" prevents pan from ever asking gmane for articles that
> I click on in the header pane.
>
> Using a packet sniffer I can see pan requesting the articles from my
> primary server, which then takes about twenty seconds to return "article
> not found".
>
> At that point shouldn't pan then request the article from gmane as a
> backup server? Do I misunderstand pan's definition of "backup"?
Hmm. I don't think you misunderstand.
What's your number of connections allowed for gmane? If it's set to
zero, then indeed, pan won't try to fetch from that server, because it
takes zero connections as disabled.
This was a feature added some time ago on request, for those who had
servers with bandwidth limits that might be exceeded, so they could
temporarily disable the server and not constantly try to download from it
when their bandwidth allowance was used up. Before that, pan would
always try to connect to any configured server -- there was no way to
temporarily disable them without deleting the server, or switching out
server.xml files between one that was enabled and another that wasn't,
without pan running.
If however you have a positive number of connections set, then pan
/should/ try the backup servers, but /only/ if an article doesn't appear
on the primary server.
BTW, while the GUI only has primary and backup, there can actually be
multiple levels set. To set additional levels, edit the servers.xml file
directly (with pan not running, of course), just as you'd do to set more
than four connections. Pan will (well, should) try the servers on one
tier first, only trying those on the next if none on the higher priority
(lower number) tiers have the post.
I had /thought/ this functionality was well established and tested, but
unless you've set the connections number to zero and are thus expecting
pan to try article fetches on a disabled server, apparently some recent
change has "disturbed the balance of the force" somewhere along the
line. Since no one else has posted about it, the first presumption would
be that it's a recent hmueller repo change. A bisect may well be in
order, possibly with a test of a khaley repo build just to nail down that
it is indeed a hmueller change, first.
But, did it download headers from gmane? It shouldn't be doing that if
connections is set to zero, and if there's at least one connection, then
it should be trying to grab posts from the backup server too.
Actually, I /think/ there's another change, too, tho I'm not /sure/ of
pan's past behavior in this regard. As I understood it, pan should only
ask for posts on groups that a server is known to carry. And all gmane
server groups should be gmane.* hierarchy and the gmane.* hierarchy
SHOULDN'T be carried by any other server, as that hierarchy is supposed
to be private and non-distributed, tho some providers may try to carry it
and get rather spotty coverage as a result. Thus, check to see if the
gmane groups are indeed carried by your other provider. If not, then I
*THINK* pan's behavior changed in that regard as well, which may be part
of the bug or a different one, tho you'd have to test either a proper
release or a khaley repo build to double-check that.
--
Duncan - List replies preferred. No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master." Richard Stallman