Thomas Jalsovsky <[EMAIL PROTECTED]> wrote: > If I'm right (I heard this nut not checked yet), NFS has hard and soft > modes. When the soft mode is turned on, every file function fail will > reported back to the application and won't try to make the job (in loop).
Yeah, but if the server is writing a lot of data, eventully soft fail will run out of buffers to store the data. > Yes, it's true. This solution should be for case when I can't store to > local disk (disk full/ local filesystem fail). My idea: > 1. try to store to local filesystem > 2. if it os OK, go next, if it FAILs, try to save to a remote FS e.g. NFS > 3. if everything FAILs (little probability) accounting returns with FAIL If you can't store to local disk, then it's likely that the whole server will die. I would suggest per-hour detail logs on local disk. Use radrelay to back them up to another machine, and delete the old ones. > p.s.: what is your idea, how should I solve filesystem problems - FS > fail-over? (I have also SQL databases, but I need for archivation those > CDRs) If all of your file system fail, then the server goes down. One option would be to have a front-end server which proxies ALL requests to a series of back-end servers. That way the front-end server has little to no local file system requirements, and the back-ends are simple and independent. When one back-end goes down, the front-end fails over to another back-end. That kind of solution is MUCH preferred to having a complicated local configuration. Alan DeKok. - List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html
