The AFS cache manager will favor a readonly volume for as long as there is a readonly path. The moment the cache manager is forced to switch to a read-write volume, it will forever access read-write volumes.
Therefore, in order to use read-only volumes there must be read-only volumes for every volume in the path starting with the root. Perhaps your problem is that: (1) you do not have readonly copies for every volume in your path (2) you are using a read-write mount point at some point in your path Jeffrey Altman Jacob Liff wrote: > Hello, > > Hrmm ok I though the way the cache manager worked was to always > favored RO copies to begin with. So what you are saying is if I create > the volume put the data into it then mount it with the .readonly > extenstion this will solve the issue I am having? > > I feel like I am still somehow missing a step here. Do I need to mount > the volume readonly on each of the servers? I didnt/dont think its > possible to have the same mountpoint name on more then one server. Feels > like Im right there just mising the final step in all this documentation. > > What I am needing to accomplish is a RO volume that will be accesible at > all times even if one of the servers suddenly dies. > > Thank you for the help, > > Jacob L. > > > -----Original Message----- > From: [EMAIL PROTECTED] on behalf of Jeffrey Altman > Sent: Fri 7/22/2005 4:48 PM > Cc: openafs-info@openafs.org > Subject: Re: [OpenAFS] replication issues 1.3.85 > > Esther Filderman wrote: >> Since RW data is static, RO data is not used to "cover" for > unavailable RW data. > > Esther meant to say: > > Since the data stored in a RW volume is volatile, a RO volume cannot be > used to provide automatic failover. If a RO volume was used to as a > failover for a RW volume that suddenly became unavailable, you would > risk suffering data corruption. > > Jeffrey Altman > > >
smime.p7s
Description: S/MIME Cryptographic Signature