remember this setup?
+--------------+ +--------------+
| | | |
| National | | National |
| Cache #1 | | Cache #2 |
| | | |
| | | |
+--------------+ +--------------+
^ parent ^ parent
| |
| Link A (DOWN) | Link B
| |
| |
+--------------+ +--------------+
| | | |
| | parent parent | |
| Cache A |<----------------------->| Cache B |
| | | |
| | | |
+--------------+ +--------------+
| |
User set A User set B
Users in group A have their cache set to Cache A. Cache A has
both National Cache #1 and Cache B as its parents.
This has been working fine.
Today, Link A went down for an extended period, and I discovered a
strange behaviour. Some pages were indeed fetched through cache B,
*but* there was a LONG delay before the pages appeared, on the order
of 20 seconds to a minute. (Unless they were already in the cache of course)
It was my understanding that there was a short timeout on the ICP requests
to the other caches to see if they already had the page, and that if one of
those caches did not respond at all, it would not even be asked for the
page, and that the other parent would be asked unconditionally to fetch
it immediately. (Which in this case should succeed because Link B and
national cache #2 are both up)
So where does this long delay come from? This was seen on ordinary
pages, and some cgi-bin pages appeared to be delivered OK, so I don't
*think* the problem is the direct fetches of pages with cgi-bin or ?
in them.
Is there some timeout somewhere that I've missed?
Or could it be something to do with the recurive relationship betwen the
two local parents?
Any help with this would be greatly appreciated. It's causing big
problems for our users.
Thanks
Graham
Received on Mon May 12 1997 - 14:02:59 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:35:10 MST