On Fri, 27 Apr 2012 14:44:25 -0400 Jeffrey Altman <[email protected]> wrote:
> Indeed it does. src/WINNT/afsd/cm_conn.c cm_Analyze() Okay, but it doesn't go through the whole list again, if I understand you correctly. So if we have two sites, A, and B, a client can think only A has new data. It contacts A, gets VOFFLINE because someone is releasing. It checks the VLDB, and sees that B is now a valid site. It contacts B, and gets VOFFLINE because now B is getting updated. The Windows client then returns an application error without trying further, correct? > However, if the volume location list does not change, the Windows > client will not reset the offline state of a down server. This is to > prevent slamming a file server with requests while the volume is > offline. Okay, but Unix clients don't fail in this situation, and as far as I know they they've had that behavior since forever (I would guess around 3.4 if I'm to believe code comments). The risk of 'slamming' a fileserver would seem low, because volumes do not stay in this state for very long, and we rate-limit ourselves from checking too often. That approach works, as far as I am aware. -- Andrew Deason [email protected] _______________________________________________ OpenAFS-info mailing list [email protected] https://lists.openafs.org/mailman/listinfo/openafs-info
