Hi, I have setup an old Compaq Proliant DL580 with 2 xeon @700MHz 2Gb RAM two SmartArray 5300 controllers and 12 drives in an array enclosure. I am running the latest opensolaris update bfu''ed from binaries since I could not build from source. I am controlling the drives with the cpqary3 driver (Solaris 10) from HP. Initially the array had 7 drives and I created a raidz zfs pool with them. The I got two more drives for the array and realized, after reading the list, that I couldn''t add them to the raidz pool. So I grabbed 3 other smaller drives and created a simple zfs pool with 5 drives to move the data from the raidz pool, in order to reconfigure it. That is, create a 4-drive raidz pool move back the data to it and then add to it a second 4-drive raidz pool. Please note that the drive with the os and the 5 "new" drives are connected to one of the 5300 controllers and the 7 drives to the other. However, when I try to move a very large file from the raidz pool to the new pool I get a reboot. I tried to run with kmdb and it shows me a page fault. The reboot occurs after a random number of Gb have been transferred to the new pool. I cannot say for sure that zfs is the culprit but it is central part in all fs operations. Also, note that I was able to create this big file through nfs (the raidz array is exported) without any problem. It is the transfer between the two zfs pools that I cannot do. Any clues or help with debugging the situation would be greatly appreciated. I am rather new to solaris (but not unix) and I do not know everything, e.g. what to do with kmdb after the page fault, to better diagnose the problem. Many thanks, George Betzos
George Betzos wrote:> Hi, > > I have setup an old Compaq Proliant DL580 with 2 xeon @700MHz 2Gb RAM > two SmartArray 5300 controllers and 12 drives in an array enclosure. > > I am running the latest opensolaris update bfu''ed from binaries since > I could not build from source. I am controlling the drives with the > cpqary3 driver (Solaris 10) from HP. > > Initially the array had 7 drives and I created a raidz zfs pool with > them. The I got two more drives for the array and realized, after > reading the list, that I couldn''t add them to the raidz pool. So I > grabbed 3 other smaller drives and created a simple zfs pool with 5 > drives to move the data from the raidz pool, in order to reconfigure > it. That is, create a 4-drive raidz pool move back the data to it and > then add to it a second 4-drive raidz pool. Please note that the drive > with the os and the 5 "new" drives are connected to one of the 5300 > controllers and the 7 drives to the other. > > However, when I try to move a very large file from the raidz pool to > the new pool I get a reboot. I tried to run with kmdb and it shows me > a page fault. The reboot occurs after a random number of Gb have been > transferred to the new pool. >What''s the exact output from kmdb? can you save a core? your console should have shown what the panic was... that''s the starting point to figure out who''s to blame... eric> I cannot say for sure that zfs is the culprit but it is central part > in all fs operations. Also, note that I was able to create this big > file through nfs (the raidz array is exported) without any problem. It > is the transfer between the two zfs pools that I cannot do. > > Any clues or help with debugging the situation would be greatly > appreciated. I am rather new to solaris (but not unix) and I do not > know everything, e.g. what to do with kmdb after the page fault, to > better diagnose the problem. > > Many thanks, > > George Betzos > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss