sdkp1-evms

Recommended update for EVMS

Hint for our patch documentation team, please work with Boyd Memmot on a really good documentation. This update has an high impact on OES2 customers. During install (and possibly at other times) there existed a window where pools could be corrupted if writing to the pool during an EVMS engine open. This was caused by a switch of which EVMS segment manager claimed the device. During the window, the write requests could potentially go to a device mapper object that was 32 sectors different than the correct position. The EVMS package was changed to modify the DOS Segment manager to not accept devices that were mapped to the NW Segment Manager. In addition, the NW Segment Manager code was modified to not accept devices that were mapped to the DOS Segment manager. The NW Segment Manager code was included in the EVMS package so that both would be installed at the same time. This avoided a problem where only one of the managers were updated, leaving a window of possible corruption. Info about the bug: After a system reboot , when nss loaded on the server , we received a crash. Adding the line /poolautodeactivate=all to the /etc/opt/novell/nss/conf/nssstart.conf and rebuild the pool nss could start normally again. While I can understand that rebuilding the pool resolved the crash , the nss should not crash the server or not start if there is corruption in a pool , we should just put the pool into maintenace mode. The core can be accessed on 147.2.77.182 under /home/core/. I have attached the analysis.log to the bug. crash> bt PID: 20395 TASK: f660ed30 CPU: 1 COMMAND: "mount" #0 [f63cd73c] crash_kexec at c013bb1a #1 [f63cd780] die at c01055fe #2 [f63cd7b0] do_page_fault at c02995e5 #3 [f63cd804] error_code (via page_fault) at c0104e4d EAX: 75c72020 EBX: 20000000 ECX: 00000000 EDX: 10000000 EBP: f5c72000 DS: 007b ESI: 00000001 ES: 007b EDI: 01464c61 CS: 0060 EIP: f9b1e9d2 ERR: ffffffff EFLAGS: 00210217 #4 [f63cd838] findIndexInXLeaf at f9b1e9d2 #5 [f63cd884] mpkPrivateFree at f8be1317 #6 [f63cd8a4] LB_StackFree at f8cd5213 #7 [f63cd8b4] ZFSVOL_LoadSystemBeasts at f9b1c016 #8 [f63cd924] ZFSPOOL_VOL_ChangeVolumeState at f9b16411 #9 [f63cd974] VOL_SingleStateChange at f90a7d45 EAX: 00000015 EBX: 08068db0 ECX: 08068dc8 EDX: 08068df0 DS: 007b ESI: c0ed0000 ES: 007b EDI: 08068e20 SS: 007b ESP: bfed1470 EBP: bfed14b8 CS: 0073 EIP: ffffe410 ERR: 00000015 EFLAGS: 00200246 Hint for our patch documentation team, please work with Boyd Memmot on a really good documentation. This update has an high impact on OES2 customers. During install (and possibly at other times) there existed a window where pools could be corrupted if writing to the pool during an EVMS engine open. This was caused by a switch of which EVMS segment manager claimed the device. During the window, the write requests could potentially go to a device mapper object that was 32 sectors different than the correct position. The EVMS package was changed to modify the DOS Segment manager to not accept devices that were mapped to the NW Segment Manager. In addition, the NW Segment Manager code was modified to not accept devices that were mapped to the DOS Segment manager. The NW Segment Manager code was included in the EVMS package so that both would be installed at the same time. This avoided a problem where only one of the managers were updated, leaving a window of possible corruption. Info about the bug: After a system reboot , when nss loaded on the server , we received a crash. Adding the line /poolautodeactivate=all to the /etc/opt/novell/nss/conf/nssstart.conf and rebuild the pool nss could start normally again. While I can understand that rebuilding the pool resolved the crash , the nss should not crash the server or not start if there is corruption in a pool , we should just put the pool into maintenace mode. The core can be accessed on 147.2.77.182 under /home/core/. I have attached the analysis.log to the bug. crash> bt PID: 20395 TASK: f660ed30 CPU: 1 COMMAND: "mount" #0 [f63cd73c] crash_kexec at c013bb1a #1 [f63cd780] die at c01055fe #2 [f63cd7b0] do_page_fault at c02995e5 #3 [f63cd804] error_code (via page_fault) at c0104e4d EAX: 75c72020 EBX: 20000000 ECX: 00000000 EDX: 10000000 EBP: f5c72000 DS: 007b ESI: 00000001 ES: 007b EDI: 01464c61 CS: 0060 EIP: f9b1e9d2 ERR: ffffffff EFLAGS: 00210217 #4 [f63cd838] findIndexInXLeaf at f9b1e9d2 #5 [f63cd884] mpkPrivateFree at f8be1317 #6 [f63cd8a4] LB_StackFree at f8cd5213 #7 [f63cd8b4] ZFSVOL_LoadSystemBeasts at f9b1c016 #8 [f63cd924] ZFSPOOL_VOL_ChangeVolumeState at f9b16411 #9 [f63cd974] VOL_SingleStateChange at f90a7d45 EAX: 00000015 EBX: 08068db0 ECX: 08068dc8 EDX: 08068df0 DS: 007b ESI: c0ed0000 ES: 007b EDI: 08068e20 SS: 007b ESP: bfed1470 EBP: bfed14b8 CS: 0073 EIP: ffffe410 ERR: 00000015 EFLAGS: 00200246 recommended evms i586 e0574145aa2819dd1850e32776e1bd934115b18c evms-gui i586 f52e21692e1c7c5e280e6c1f7ea2d5a10e785d95 evms-ha i586 d8039c2acf41266b469379b7b2a56d9d5662e2d5