On Sunday July 8, [EMAIL PROTECTED] wrote: > > Where I'm at now: > > I am using Mandrake 8.0 (kernel 2.4.3). Not much has changed since 2.4.3. A number of code improvements that don't exhibit any function difference. A bug fix in 2.4.5 which isn't related. A number of substantial performance improvements for raid5 went into 2.4.6, so you might consider upgrading to that release. > > > The problem: > > When I try to create a filesystem on the degraded array, or when the > 3-disk array tries to synchronise, I get the following message: > > raid 5: switching cache buffer size, 4096 --> 1024 This is a benign informational message. It is only a concern if you get a lot of them. raid5 maintains a stripe cache which must be the same size as the IO requests that it is receiving. It default to 4096 until the first request. mkfs and fsck use a block size of 1024, hence the switch. Once you mount the filesystem, it will switch back to 4096 (assuming a 4K block size of the filesystem which is normal). When you get this message, it means that a request of a different size has arrived, so raid5 flushes its cache and then rebuilds it with the new size. Someone suggested that such messages should be followed by a "size now" message. This isn't true. The "size now" is printed with a "PRINTK", not a "printk", and is only actually printed if debugging is compiled in. > > and the system freezes, requiring a hard reset. Is this "and then immediately the system freezes", or it is "and then eventually the system freezes, but there are no more messages"? Does mkfs report any progress at all? Are you using a kernel that you compiled yourself, or is it a prepackaged kernel from Mandrake. If the later, then it undoubtedly includes a few extra patches. I doubt they would have broken raid, but anything is possible. I suggest you compile a clear 2.4.6 kernel and see what happens then. Also, you could try "alt-sysrq-t" to get a task listing when it has hung, and see what is happening ... though capturing that and mailing it might be awkward. Running mkfs on a degraded array certainly works for me on my hardware. NeilBrown - To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to [EMAIL PROTECTED]
