squinn@ncsa.uiuc.edu (Steve Quinn) (04/03/91)
We recently moved our Sybase installation from Sun3/SunOS 4.0/SQLServer 4.0.1 to Sun4/SunOS4.1.1/SQLServer 4.0.1. We had previously been doing our database dumps to an xabyte drive, doing multiple dumps to one xabyte tape. Since our move "up" in hardware and OS, I haven't been able to get this to work. There were actually two problems with this; one produced a message something like "dbsopen() failed blah blah blah" but Sybase sent me an EBF for this whiched worked. However, the other problem didn't get fixed. Whatever this problem is, it ultimatly produces the following message on the console: "vmunix: tape synchronization lost rewinding tape" or something very much like that. Sybase does not actually support multi volume dumps to one tape, so they are giving little or no help with this problem. Anyone have any ideas on this? An additional note to this: ANOTHER problem we had in this harware/software migration was that the server will mysteriously go into a permanent resource wait periodically. This process shows status D when you run ps. This is a real pain, since processes with status D cannot be killed, so you need to reboot the machine. Occasionally this will actually hang up the machine, such that you need to physically power off and on the cpu (l1a won't even work). Sybase claims this is due to a bug in SunOS 4.1.1, but only on Sun4/260s and 280s. They claim that a combination of backing up to OS 4.1, installing Sun's Database Accelerator, and installing an OS patch or two will fix this problem (nothing to it, right?). It is also not clear from the info from either Sun or Sybase that all three of these steps are necessary. I've heard a rumor from Sybase that there is a possible work around for the first bug I mentioned above (you remember, the xabyte bug) but the rumour has it that this work around is only for OS 4.1.1. Assuming this rumour is true, I'm stuck with either staying at OS 4.1.1 and (possibly) working around the xabyte problem, while living with periodic system hangs requiring reboots, OR backing off from 4.1.1, "probably" taking care of the system hang problem, but being unable to use our current dumping scheme. The lesson to be learned from all of this is: NEVER NEVER UPGRADE ANYTHING. If anyone has seen/heard any solutions to any of these problems, please respond or email me directly. I'll be happy to summarize any responses for those poor soles needing to do these upgrades in the future.