Re: [OpenAFS] /afs area is hanging
> If you wipe out your cache, AFS does tend to perform badly. I > recommend against it ;) Could afsd something like chmod 0 or use some more fancy stuff to prevent userland stuff to get at the cache? Harald. ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info
Re: [OpenAFS] /afs area is hanging
Mark Henry wrote: This is a happy day. Derrick it looks like your skills have pointed us to the solution. I mentioned that we use a loopback device to mount our afs cache filesystem. That device was /dev/loop0. Well, after the direction that you gave us we found that /dev/loop0 was also being used as a method of restricting font cache for a different app. When the app would run the afs cache was getting clobbered and the afs hang would follow. We have moved the afs cache to a new place now and it looks like this problem has been solved. Thank you all on openafs.org that helped us with this issue. Thank you Derrick for the key piece of info that has solved this one. I'm curious what the backing store for /dev/loop0 is in your setup. what advantages do you receive while running this way? Is this so you can store the cache in a ramdisk? Thanks, Jason ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info
Re: [OpenAFS] /afs area is hanging
On Wed, May 13, 2009 at 11:16 AM, Mark Henry wrote: >> that may well be related to the problem. what cron jobs do you have? > >> Derrick > > This is a happy day. Derrick it looks like your skills have pointed us to the > solution. I mentioned that we use a loopback device to mount our afs cache > filesystem. That device was /dev/loop0. Well, after the direction that you > gave us we found that /dev/loop0 was also being used as a method of > restricting > font cache for a different app. When the app would run the afs cache was > getting clobbered and the afs hang would follow. We have moved the afs cache > to a new place now and it looks like this problem has been solved. Thank you > all on openafs.org that helped us with this issue. Thank you Derrick for the > key piece of info that has solved this one. If you wipe out your cache, AFS does tend to perform badly. I recommend against it ;) That said, we can (and possibly also will) improve the behavior a bit. Derrick ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info
Re: [OpenAFS] /afs area is hanging
> that may well be related to the problem. what cron jobs do you have? > Derrick This is a happy day. Derrick it looks like your skills have pointed us to the solution. I mentioned that we use a loopback device to mount our afs cache filesystem. That device was /dev/loop0. Well, after the direction that you gave us we found that /dev/loop0 was also being used as a method of restricting font cache for a different app. When the app would run the afs cache was getting clobbered and the afs hang would follow. We have moved the afs cache to a new place now and it looks like this problem has been solved. Thank you all on openafs.org that helped us with this issue. Thank you Derrick for the key piece of info that has solved this one. On May 12, 2009, at 6:25 PM, Mark Henry wrote: >>> I take it it;'s not likely that something is deleteing files out from >>> under AFS inside this filesystem? >> >> Not that I am aware of. We did notice that some of the files were >> missing in >> one of the system's cache areas. One of the systems does not have the >> following files: CacheItems, CellItems, VolumeItems and the dir lost >> +found. >> >> Mark Henry Mark Henry _ "This message and any attachments are solely for the intended recipient and may contain confidential or privileged information. If you are not the intended recipient, any disclosure, copying, use, or distribution of the information included in this message and any attachments is prohibited. If you have received this communication in error, please notify us by reply e-mail and immediately and permanently delete this message and any attachments. Thank you." _ ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info
Re: [OpenAFS] /afs area is hanging
that may well be related to the problem. what cron jobs do you have? Derrick On May 12, 2009, at 6:25 PM, Mark Henry wrote: I take it it;'s not likely that something is deleteing files out from under AFS inside this filesystem? Not that I am aware of. We did notice that some of the files were missing in one of the system's cache areas. One of the systems does not have the following files: CacheItems, CellItems, VolumeItems and the dir lost +found. Mark Henry _ "This message and any attachments are solely for the intended recipient and may contain confidential or privileged information. If you are not the intended recipient, any disclosure, copying, use, or distribution of the information included in this message and any attachments is prohibited. If you have received this communication in error, please notify us by reply e-mail and immediately and permanently delete this message and any attachments. Thank you." _ ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info
Re: [OpenAFS] /afs area is hanging
> I take it it;'s not likely that something is deleteing files out from > under AFS inside this filesystem? Not that I am aware of. We did notice that some of the files were missing in one of the system's cache areas. One of the systems does not have the following files: CacheItems, CellItems, VolumeItems and the dir lost+found. Mark Henry _ "This message and any attachments are solely for the intended recipient and may contain confidential or privileged information. If you are not the intended recipient, any disclosure, copying, use, or distribution of the information included in this message and any attachments is prohibited. If you have received this communication in error, please notify us by reply e-mail and immediately and permanently delete this message and any attachments. Thank you." _ ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info
Re: [OpenAFS] /afs area is hanging
On Tue, May 12, 2009 at 6:05 PM, Mark Henry wrote: > >> yes, that's very interesting. you've probably told us what we need to. >> you got an oops in the thread holding the lock, the machine will never >> recover. what filesystem is behind your afs cache? > > We use a virtual ext3 filesystem mounted on a loopback device. We have done > this with many afs clients and it seems to be working. Here is our fstab > entry: > > /AFSvirtualFS /usr/vice/cache ext3 defaults,loop=/dev/loop0 0 0 I take it it;'s not likely that something is deleteing files out from under AFS inside this filesystem? ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info
Re: [OpenAFS] /afs area is hanging
> yes, that's very interesting. you've probably told us what we need to. > you got an oops in the thread holding the lock, the machine will never > recover. what filesystem is behind your afs cache? We use a virtual ext3 filesystem mounted on a loopback device. We have done this with many afs clients and it seems to be working. Here is our fstab entry: /AFSvirtualFS /usr/vice/cache ext3 defaults,loop=/dev/loop0 0 0 Mark Henry Derrick Brashear 05/12/2009 02:58 PM To Mark Henry cc Felix Frank , openafs-info@openafs.org, Simon Wilkinson Subject Re: [OpenAFS] /afs area is hanging On Tue, May 12, 2009 at 4:37 PM, Mark Henry wrote: > > > >> I took the liberty to paste the interesting parts to >> http://pastebin.com/m53578cd5. Notice the bottom, which was the original >> bottom as well. Mark, you've been asked to look at dmesg before this, so I >> suppose this didn't happen before you tried this call-tracing? > >> Besides, it would be interesting if an upgrade to 1.4.10 makes the problem >> go away. Can you try that? > > We upgraded to 1.4.10. We got the same errors. Here is the cmdebug output: > > => cmdebug HOSTNAME > ** Cache entry @ 0x172d1880 for 2.536870937.28.1820 [afs.dev.infoprint.com] > locks: (none_waiting, write_locked(pid:-246839824 at:681)) >3 bytes DV0 refcnt 2 > callback 22174d40 expires 1242158668 > 0 opens 0 writers > mount point > states (0x5), stat'd, read-only which is afs_HandleLink, called from afs_lookup. Memcache or disk cache? (different implementations of that function) > > The volume # 536870937 above just happens to be root.cell but it has been > different based on which dir I do the ls in. > >> Finally, it looks like ls and sshd got locked up trying to determine your >> client machine's home cell. I can see that happening (only) if no cell has >> been set at that point. The output of fs wscell would be interesting in >> this situation, but I'm not sure wether that would lock up as well (and if >> it is at all helpful). > > We tried the fs wscell command. It worked fine if the fs command was local > and hung if the fs was being retrieved from afs. well, if afs is unhappy, one presumes afs is unhappy. > Also, here is a bit of interesting output from dmesg when the system is > hung: ah, that'd be disk cache. > AssertProcessEntry: pohm_main, pid=6518 > openafs: Can't open inode 95550 yes, that's very interesting. you've probably told us what we need to. you got an oops in the thread holding the lock, the machine will never recover. what filesystem is behind your afs cache? > [ cut here ] > kernel BUG at > /compile/openafs-1.4.10/src/libafs/MODLOAD-2.6.22.5-31-default-MP/osi_file.c:87! > invalid opcode: [1] SMP > last sysfs file: /class/scsi_host/host0/model > CPU 6 > Modules linked in: tun iptable_filter ip_tables x_tables amk ipv6 libafs(P) > microcode firmware_class usbhid hid ff_memless tp_dd af_packet apparmor ext2 > loop dm_mod parport_pc parport bnx2 rtc_cmos rtc_core i2c_i801 rtc_lib > ide_cd i2c_core cdrom shpchp tg3 pci_hotplug container button sg ehci_hcd > uhci_hcd usbcore sd_mod ata_piix libata edd ext3 mbcache jbd fan aacraid > scsi_mod piix ide_core thermal processor > Pid: 6519, comm: ASCIIMast Tainted: P N 2.6.22.5-31-default #1 > RIP: 0010:[] [] > :libafs:osi_UFSOpen+0x155/0x1f2 > RSP: :81031de299c8 EFLAGS: 00010296 > RAX: 0023 RBX: 81031a92b418 RCX: 0001 > RDX: 804bdfe8 RSI: 0096 RDI: 804bdfe0 > RBP: 81042145f000 R08: 0001 R09: 810001086bc0 > R10: 0046 R11: 81042fa74ec0 R12: 81042af37000 > R13: 0001753e R14: 0001753e R15: 0003 > FS: () GS:81042ee9f1c0() knlGS: > CS: 0010 DS: 002b ES: 002b CR0: 8005003b > CR2: f5d6050c CR3: 00039e0b9000 CR4: 06e0 > Process ASCIIMast (pid: 6519, threadinfo 81031de28000, task > 81031ad770c0) > Stack: c2000770eb90 c2000770caa8 81042a631000 8104172d1880 > 81039800839c 8832f678 0003 0003 > c2000770eb90 8104172d1880 > Call Trace: > [] :libafs:afs_UFSHandleLink+0xf7/0x1bd > [] :libafs:afs_lookup+0xbb2/0x115f > [] :libafs:afs_linux_dentry_revalidate+0x422/0x434 > [] :libafs:afs_linux_lookup+0x85/0x1ca > [] :libafs:PagInCred+0x30/0xa9 > [] do_lookup+0xc4/0x1ae > [] __link_path_walk+0x36c/0xd8b > [] dput+0x26/0x115 > [] __link_path_walk+0xc38/0xd8b >
Re: [OpenAFS] /afs area is hanging
On Tue, May 12, 2009 at 4:37 PM, Mark Henry wrote: > > > >> I took the liberty to paste the interesting parts to >> http://pastebin.com/m53578cd5. Notice the bottom, which was the original >> bottom as well. Mark, you've been asked to look at dmesg before this, so I >> suppose this didn't happen before you tried this call-tracing? > >> Besides, it would be interesting if an upgrade to 1.4.10 makes the problem >> go away. Can you try that? > > We upgraded to 1.4.10. We got the same errors. Here is the cmdebug output: > > => cmdebug HOSTNAME > ** Cache entry @ 0x172d1880 for 2.536870937.28.1820 [afs.dev.infoprint.com] > locks: (none_waiting, write_locked(pid:-246839824 at:681)) > 3 bytes DV 0 refcnt 2 > callback 22174d40 expires 1242158668 > 0 opens 0 writers > mount point > states (0x5), stat'd, read-only which is afs_HandleLink, called from afs_lookup. Memcache or disk cache? (different implementations of that function) > > The volume # 536870937 above just happens to be root.cell but it has been > different based on which dir I do the ls in. > >> Finally, it looks like ls and sshd got locked up trying to determine your >> client machine's home cell. I can see that happening (only) if no cell has >> been set at that point. The output of fs wscell would be interesting in >> this situation, but I'm not sure wether that would lock up as well (and if >> it is at all helpful). > > We tried the fs wscell command. It worked fine if the fs command was local > and hung if the fs was being retrieved from afs. well, if afs is unhappy, one presumes afs is unhappy. > Also, here is a bit of interesting output from dmesg when the system is > hung: ah, that'd be disk cache. > AssertProcessEntry: pohm_main, pid=6518 > openafs: Can't open inode 95550 yes, that's very interesting. you've probably told us what we need to. you got an oops in the thread holding the lock, the machine will never recover. what filesystem is behind your afs cache? > [ cut here ] > kernel BUG at > /compile/openafs-1.4.10/src/libafs/MODLOAD-2.6.22.5-31-default-MP/osi_file.c:87! > invalid opcode: [1] SMP > last sysfs file: /class/scsi_host/host0/model > CPU 6 > Modules linked in: tun iptable_filter ip_tables x_tables amk ipv6 libafs(P) > microcode firmware_class usbhid hid ff_memless tp_dd af_packet apparmor ext2 > loop dm_mod parport_pc parport bnx2 rtc_cmos rtc_core i2c_i801 rtc_lib > ide_cd i2c_core cdrom shpchp tg3 pci_hotplug container button sg ehci_hcd > uhci_hcd usbcore sd_mod ata_piix libata edd ext3 mbcache jbd fan aacraid > scsi_mod piix ide_core thermal processor > Pid: 6519, comm: ASCIIMast Tainted: P N 2.6.22.5-31-default #1 > RIP: 0010:[] [] > :libafs:osi_UFSOpen+0x155/0x1f2 > RSP: :81031de299c8 EFLAGS: 00010296 > RAX: 0023 RBX: 81031a92b418 RCX: 0001 > RDX: 804bdfe8 RSI: 0096 RDI: 804bdfe0 > RBP: 81042145f000 R08: 0001 R09: 810001086bc0 > R10: 0046 R11: 81042fa74ec0 R12: 81042af37000 > R13: 0001753e R14: 0001753e R15: 0003 > FS: () GS:81042ee9f1c0() knlGS: > CS: 0010 DS: 002b ES: 002b CR0: 8005003b > CR2: f5d6050c CR3: 00039e0b9000 CR4: 06e0 > Process ASCIIMast (pid: 6519, threadinfo 81031de28000, task > 81031ad770c0) > Stack: c2000770eb90 c2000770caa8 81042a631000 8104172d1880 > 81039800839c 8832f678 0003 0003 > c2000770eb90 8104172d1880 > Call Trace: > [] :libafs:afs_UFSHandleLink+0xf7/0x1bd > [] :libafs:afs_lookup+0xbb2/0x115f > [] :libafs:afs_linux_dentry_revalidate+0x422/0x434 > [] :libafs:afs_linux_lookup+0x85/0x1ca > [] :libafs:PagInCred+0x30/0xa9 > [] do_lookup+0xc4/0x1ae > [] __link_path_walk+0x36c/0xd8b > [] dput+0x26/0x115 > [] __link_path_walk+0xc38/0xd8b > [] link_path_walk+0x58/0xe0 > [] do_filp_open+0x1c/0x3d > [] do_path_lookup+0x1ab/0x227 > [] __path_lookup_intent_open+0x56/0x97 > [] open_namei+0x7a/0x674 > [] do_filp_open+0x1c/0x3d > [] do_sys_open+0x44/0xc1 > [] ia32_sysret+0x0/0xa > > > We still can't seem to get this system to stop hanging in the /afs area. > > Mark Henry > > > _ > "This message and any attachments are solely for the intended recipient and > may contain confidential or privileged information. If you are not the > intended recipient, any disclosure, copying, use, or distribution of the > information included in this message and any attachments is prohibited. If > you have received this communication in error, please notify us by reply > e-mail and immediately and permanently delete this message and any > attachments. Thank you." > _ > --
Re: [OpenAFS] /afs area is hanging
> I took the liberty to paste the interesting parts to > http://pastebin.com/m53578cd5. Notice the bottom, which was the original > bottom as well. Mark, you've been asked to look at dmesg before this, so I > suppose this didn't happen before you tried this call-tracing? > Besides, it would be interesting if an upgrade to 1.4.10 makes the problem > go away. Can you try that? We upgraded to 1.4.10. We got the same errors. Here is the cmdebug output: => cmdebug HOSTNAME ** Cache entry @ 0x172d1880 for 2.536870937.28.1820 [afs.dev.infoprint.com] locks: (none_waiting, write_locked(pid:-246839824 at:681)) 3 bytes DV0 refcnt 2 callback 22174d40 expires 1242158668 0 opens 0 writers mount point states (0x5), stat'd, read-only The volume # 536870937 above just happens to be root.cell but it has been different based on which dir I do the ls in. > Finally, it looks like ls and sshd got locked up trying to determine your > client machine's home cell. I can see that happening (only) if no cell has > been set at that point. The output of fs wscell would be interesting in > this situation, but I'm not sure wether that would lock up as well (and if > it is at all helpful). We tried the fs wscell command. It worked fine if the fs command was local and hung if the fs was being retrieved from afs. Also, here is a bit of interesting output from dmesg when the system is hung: AssertProcessEntry: pohm_main, pid=6518 openafs: Can't open inode 95550 [ cut here ] kernel BUG at /compile/openafs-1.4.10/src/libafs/MODLOAD-2.6.22.5-31-default-MP/osi_file.c:87! invalid opcode: [1] SMP last sysfs file: /class/scsi_host/host0/model CPU 6 Modules linked in: tun iptable_filter ip_tables x_tables amk ipv6 libafs(P) microcode firmware_class usbhid hid ff_memless tp_dd af_packet apparmor ext2 loop dm_mod parport_pc parport bnx2 rtc_cmos rtc_core i2c_i801 rtc_lib ide_cd i2c_core cdrom shpchp tg3 pci_hotplug container button sg ehci_hcd uhci_hcd usbcore sd_mod ata_piix libata edd ext3 mbcache jbd fan aacraid scsi_mod piix ide_core thermal processor Pid: 6519, comm: ASCIIMast Tainted: P N 2.6.22.5-31-default #1 RIP: 0010:[] [] :libafs:osi_UFSOpen+0x155/0x1f2 RSP: :81031de299c8 EFLAGS: 00010296 RAX: 0023 RBX: 81031a92b418 RCX: 0001 RDX: 804bdfe8 RSI: 0096 RDI: 804bdfe0 RBP: 81042145f000 R08: 0001 R09: 810001086bc0 R10: 0046 R11: 81042fa74ec0 R12: 81042af37000 R13: 0001753e R14: 0001753e R15: 0003 FS: () GS:81042ee9f1c0() knlGS: CS: 0010 DS: 002b ES: 002b CR0: 8005003b CR2: f5d6050c CR3: 00039e0b9000 CR4: 06e0 Process ASCIIMast (pid: 6519, threadinfo 81031de28000, task 81031ad770c0) Stack: c2000770eb90 c2000770caa8 81042a631000 8104172d1880 81039800839c 8832f678 0003 0003 c2000770eb90 8104172d1880 Call Trace: [] :libafs:afs_UFSHandleLink+0xf7/0x1bd [] :libafs:afs_lookup+0xbb2/0x115f [] :libafs:afs_linux_dentry_revalidate+0x422/0x434 [] :libafs:afs_linux_lookup+0x85/0x1ca [] :libafs:PagInCred+0x30/0xa9 [] do_lookup+0xc4/0x1ae [] __link_path_walk+0x36c/0xd8b [] dput+0x26/0x115 [] __link_path_walk+0xc38/0xd8b [] link_path_walk+0x58/0xe0 [] do_filp_open+0x1c/0x3d [] do_path_lookup+0x1ab/0x227 [] __path_lookup_intent_open+0x56/0x97 [] open_namei+0x7a/0x674 [] do_filp_open+0x1c/0x3d [] do_sys_open+0x44/0xc1 [] ia32_sysret+0x0/0xa We still can't seem to get this system to stop hanging in the /afs area. Mark Henry _ "This message and any attachments are solely for the intended recipient and may contain confidential or privileged information. If you are not the intended recipient, any disclosure, copying, use, or distribution of the information included in this message and any attachments is prohibited. If you have received this communication in error, please notify us by reply e-mail and immediately and permanently delete this message and any attachments. Thank you." _
Re: [OpenAFS] /afs area is hanging
On Thu, 7 May 2009, Ted Creedon wrote: Typical for 3 servers all of which hung after "ls /afs" I niced affsd to get enough keyboard control to see that klogd and syslog-ng are using all the cycles filling /var/log/messages with rx carps. I think that a massive error caused by mis-keying should not hang windows and linux clients... Must be a different problem then, Mark reported no suspicious log entries whatsoever. On Thu, 7 May 2009, Simon Wilkinson wrote: On 7 May 2009, at 21:51, Mark Henry wrote: I ran the 'echo t' command recommended below once afs hung again. It definitely put some output in dmesg. The only D states that were listed were bash sessions (I think). It looks like they were sessions that I opened after the user told me that afs was hung again. I tried to attach the dmesg output and cmdebug output but the email was rejected because the log files were way too big. Any ideas of what to try next? Or is there anything in particular that I should look at in the cmdebug or dmesg output. Thanks, Put the files somewhere we can fetch them - either on the web, or in AFS. Your cmdebug output really shouldn't be that long, though (you want to run 'cmdebug' _not_ 'cmdebug -long) My CC came through fine. The dmesg output looks mangled, however, it starts in the middle of a trace. I hope nothing vital is missing, but I failed to find afsd, which was mildly disappointing. I took the liberty to paste the interesting parts to http://pastebin.com/m53578cd5. Notice the bottom, which was the original bottom as well. Mark, you've been asked to look at dmesg before this, so I suppose this didn't happen before you tried this call-tracing? Besides, it would be interesting if an upgrade to 1.4.10 makes the problem go away. Can you try that? Finally, it looks like ls and sshd got locked up trying to determine your client machine's home cell. I can see that happening (only) if no cell has been set at that point. The output of fs wscell would be interesting in this situation, but I'm not sure wether that would lock up as well (and if it is at all helpful). Cheers - Felixsshd S 61bfbfdb6613 0 8543 1 (NOTLB) 8103489399d8 0086 8830848d 810348939988 80617800 80617800 8061d210 80617800 80617800 883082f3 Call Trace: [] :libafs:afs_IsPrimaryCell+0x2c/0x36 [] :libafs:afs_GetCellStale+0x42/0x4b [] :libafs:afs_osi_SleepSig+0xc5/0x161 [] default_wake_function+0x0/0xe [] :libafs:EvalMountPoint+0x2fa/0x3d1 [] :libafs:afs_osi_Sleep+0x69/0xba [] :libafs:Afs_Lock_Obtain+0xba/0x1a2 [] :libafs:afs_lookup+0xe67/0x115f [] :libafs:afs_pag_match+0x0/0x17 [] :libafs:afs_linux_dentry_revalidate+0x178/0x434 [] igrab+0x25/0x34 [] :libafs:afs_FindVCache+0x222/0x4f5 [] :libafs:afs_AccessOK+0x4a/0x14c [] mutex_lock+0xd/0x1e [] iput+0x42/0x7b [] :libafs:afs_PutVCache+0x93/0x10d [] :libafs:afs_PutFakeStat+0x3d/0x42 [] :libafs:afs_access+0x382/0x3be [] __d_lookup+0xbc/0x10e [] do_lookup+0x157/0x1ae [] __link_path_walk+0x36c/0xd8b [] link_path_walk+0x58/0xe0 [] audit_syscall_entry+0x138/0x17a [] do_path_lookup+0x1ab/0x227 [] getname+0x14c/0x1af [] __user_walk_fd+0x37/0x4c [] sys_faccessat+0x9c/0x148 [] tracesys+0xdc/0xe1 lsS 61c6b4bd29d2 0 8960 1 (NOTLB) 8103f4e23968 0082 8830848d 8103f4e23918 80617800 80617800 8061d210 80617800 80617800 883082f3 Call Trace: [] :libafs:afs_IsPrimaryCell+0x2c/0x36 [] :libafs:afs_GetCellStale+0x42/0x4b [] :libafs:afs_osi_SleepSig+0xc5/0x161 [] default_wake_function+0x0/0xe [] :libafs:EvalMountPoint+0x2fa/0x3d1 [] :libafs:afs_osi_Sleep+0x69/0xba [] :libafs:Afs_Lock_Obtain+0xba/0x1a2 [] :libafs:afs_lookup+0xe67/0x115f [] :libafs:afs_AccessOK+0x4a/0x14c [] :libafs:afs_linux_dentry_revalidate+0x178/0x434 [] igrab+0x25/0x34 [] :libafs:afs_FindVCache+0x222/0x4f5 [] :libafs:afs_AccessOK+0x4a/0x14c [] mutex_lock+0xd/0x1e [] iput+0x42/0x7b [] :libafs:afs_PutVCache+0x93/0x10d [] :libafs:afs_PutFakeStat+0x3d/0x42 [] :libafs:afs_access+0x382/0x3be [] __d_lookup+0xbc/0x10e [] do_lookup+0x157/0x1ae [] __link_path_walk+0x8d9/0xd8b [] :libafs:afs_TraverseCells+0x4a/0x9b [] link_path_walk+0x58/0xe0 [] current_fs_time+0x3b/0x40 [] __mark_inode_dirty+0xdf/0x17c [] do_path_lookup+0x1ab/0x227 [] getname+0x14c/0x1af [] __user_walk_fd+0x37/0x4c [] vfs_lstat_fd+0x18/0x47 [] __mark_inode_dirty+0xdf/0x17c [] sys_newlstat+0x19/0x31 [] tracesys+0x71/0xe1 [] tracesys+0xdc/0xe1 bash D 61d03f4a273d 0 9098 9096 (NOTLB) 81035ddbfe78 0082 8832198d 810341013c9c 81035ddbfe28 80617800 80617800 8061d210 80617800 80617800 8102e3f31780 Call Trace: [] :libafs:afs_getatt
Fwd: [OpenAFS] /afs area is hanging
-- Forwarded message -- From: Ted Creedon Date: Thu, May 7, 2009 at 2:57 PM Subject: Re: [OpenAFS] /afs area is hanging To: Mark Henry Use select cut and paste typical portions of dmesg and /var/log/messages I fixed my problem by recreating all the principals using "kadmin.local -e des-cbc-crc:normal" which forces all the keys to be single des. Typical for 3 servers all of which hung after "ls /afs" I niced affsd to get enough keyboard control to see that klogd and syslog-ng are using all the cycles filling /var/log/messages with rx carps. I think that a massive error caused by mis-keying should not hang windows and linux clients... A key check diagnostic would certainly help. 1.5.59 win and 1.4.10 linux on 3 suse 10.2 and 11.1 server boxes, one is dual homed Seems to work fine now. On Thu, May 7, 2009 at 1:51 PM, Mark Henry wrote: > > I ran the 'echo t' command recommended below once afs hung again. It > definitely put some output in dmesg. The only D states that were listed > were bash sessions (I think). It looks like they were sessions that I > opened after the user told me that afs was hung again. I tried to attach > the dmesg output and cmdebug output but the email was rejected because the > log files were way too big. Any ideas of what to try next? Or is there > anything in particular that I should look at in the cmdebug or dmesg output. > Thanks, > > Mark Henry > > > > *Felix Frank * > > 05/04/2009 11:46 PM > To > Mark Henry cc > openafs-info@openafs.org Subject > Re: [OpenAFS] /afs area is hanging > > > > > On Mon, 4 May 2009, Mark Henry wrote: > > > I tried the -fakestat-all option and it did not work. > > Weird. > > > I have searched /var/log/messages. I have checked the config files. > User > > authentication works fine (even when the system is hanging). If I run > the > > command 'ls -l /afs' that window is hung (or any other command that > > references afs). If an afs user logs in when the system is in a bad > state > > the session immediately hangs because it can't cd to the afs home dir. I > > don't know what to do other than reboot. Can someone tell me what else > to > > try to find out why this system is hanging? Thanks, > > You can find out just which call gets stuck by issuing an > 'echo t >>/proc/sysrq-trigger'. Call traces for all processes can then be > found in dmesg. The broken processes are probably in a D state. > > This will probably not identify the root cause, but may give a clue about > what's going on. > > HTH > - Felix > > > > _ > "This message and any attachments are solely for the intended recipient and > may contain confidential or privileged information. If you are not the > intended recipient, any disclosure, copying, use, or distribution of the > information included in this message and any attachments is prohibited. If > you have received this communication in error, please notify us by reply > e-mail and immediately and permanently delete this message and any > attachments. Thank you." > _ >
Re: [OpenAFS] /afs area is hanging
On 7 May 2009, at 21:51, Mark Henry wrote: I ran the 'echo t' command recommended below once afs hung again. It definitely put some output in dmesg. The only D states that were listed were bash sessions (I think). It looks like they were sessions that I opened after the user told me that afs was hung again. I tried to attach the dmesg output and cmdebug output but the email was rejected because the log files were way too big. Any ideas of what to try next? Or is there anything in particular that I should look at in the cmdebug or dmesg output. Thanks, Put the files somewhere we can fetch them - either on the web, or in AFS. Your cmdebug output really shouldn't be that long, though (you want to run 'cmdebug' _not_ 'cmdebug -long) S. ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info
Re: [OpenAFS] /afs area is hanging
I ran the 'echo t' command recommended below once afs hung again. It definitely put some output in dmesg. The only D states that were listed were bash sessions (I think). It looks like they were sessions that I opened after the user told me that afs was hung again. I tried to attach the dmesg output and cmdebug output but the email was rejected because the log files were way too big. Any ideas of what to try next? Or is there anything in particular that I should look at in the cmdebug or dmesg output. Thanks, Mark Henry Felix Frank 05/04/2009 11:46 PM To Mark Henry cc openafs-info@openafs.org Subject Re: [OpenAFS] /afs area is hanging On Mon, 4 May 2009, Mark Henry wrote: > I tried the -fakestat-all option and it did not work. Weird. > I have searched /var/log/messages. I have checked the config files. User > authentication works fine (even when the system is hanging). If I run the > command 'ls -l /afs' that window is hung (or any other command that > references afs). If an afs user logs in when the system is in a bad state > the session immediately hangs because it can't cd to the afs home dir. I > don't know what to do other than reboot. Can someone tell me what else to > try to find out why this system is hanging? Thanks, You can find out just which call gets stuck by issuing an 'echo t >>/proc/sysrq-trigger'. Call traces for all processes can then be found in dmesg. The broken processes are probably in a D state. This will probably not identify the root cause, but may give a clue about what's going on. HTH - Felix _ "This message and any attachments are solely for the intended recipient and may contain confidential or privileged information. If you are not the intended recipient, any disclosure, copying, use, or distribution of the information included in this message and any attachments is prohibited. If you have received this communication in error, please notify us by reply e-mail and immediately and permanently delete this message and any attachments. Thank you." _
Re: [OpenAFS] /afs area is hanging
Felix, Thanks for the reply. I am waiting for it to crash again to run the echo . command below. There also seems to be a separate issue with the LDAP server timing out. These messages show up in /var/log/messages. It makes the afs area really slow but not totally hung. I then restart nscd and the errors stop and all is well again. This seems to be a seperate issue from the permanent afs hang taking place. Mark Henry Felix Frank 05/04/2009 11:46 PM To Mark Henry cc openafs-info@openafs.org Subject Re: [OpenAFS] /afs area is hanging On Mon, 4 May 2009, Mark Henry wrote: > I tried the -fakestat-all option and it did not work. Weird. > I have searched /var/log/messages. I have checked the config files. User > authentication works fine (even when the system is hanging). If I run the > command 'ls -l /afs' that window is hung (or any other command that > references afs). If an afs user logs in when the system is in a bad state > the session immediately hangs because it can't cd to the afs home dir. I > don't know what to do other than reboot. Can someone tell me what else to > try to find out why this system is hanging? Thanks, You can find out just which call gets stuck by issuing an 'echo t >>/proc/sysrq-trigger'. Call traces for all processes can then be found in dmesg. The broken processes are probably in a D state. This will probably not identify the root cause, but may give a clue about what's going on. HTH - Felix _ "This message and any attachments are solely for the intended recipient and may contain confidential or privileged information. If you are not the intended recipient, any disclosure, copying, use, or distribution of the information included in this message and any attachments is prohibited. If you have received this communication in error, please notify us by reply e-mail and immediately and permanently delete this message and any attachments. Thank you." _
Re: [OpenAFS] /afs area is hanging
On Mon, 4 May 2009, Mark Henry wrote: I tried the -fakestat-all option and it did not work. Weird. I have searched /var/log/messages. I have checked the config files. User authentication works fine (even when the system is hanging). If I run the command 'ls -l /afs' that window is hung (or any other command that references afs). If an afs user logs in when the system is in a bad state the session immediately hangs because it can't cd to the afs home dir. I don't know what to do other than reboot. Can someone tell me what else to try to find out why this system is hanging? Thanks, You can find out just which call gets stuck by issuing an 'echo t >>/proc/sysrq-trigger'. Call traces for all processes can then be found in dmesg. The broken processes are probably in a D state. This will probably not identify the root cause, but may give a clue about what's going on. HTH - Felix ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info
Re: [OpenAFS] /afs area is hanging
Also, I will try to get the cmdebug output just in case there is something in there beyond what I can see. Mark Henry Jeffrey Altman 04/29/2009 11:22 AM Please respond to jalt...@secure-endpoints.com To Mark Henry cc openafs-info@openafs.org Subject Re: [OpenAFS] /afs area is hanging Mark Henry wrote: > Thank you all for your responses. I am trying the -fakestat-all option on the > afsd daemon. We will see if it works. Everything works fine for awhile after > reboot and then it is just a matter of time before everything that touches afs > hangs. Hopefully this -fakestat-all option helps. > > Mark When the cache manager hangs, execute "cmdebug " and send the output. _ "This message and any attachments are solely for the intended recipient and may contain confidential or privileged information. If you are not the intended recipient, any disclosure, copying, use, or distribution of the information included in this message and any attachments is prohibited. If you have received this communication in error, please notify us by reply e-mail and immediately and permanently delete this message and any attachments. Thank you." _
Re: [OpenAFS] /afs area is hanging
I was not able to get the cmdebug output this last time the system crashed because the user rebooted the system. We have tried running cmdebug several times when it is in its hung state and all entries say none_waiting or similar. The output looks the same as a working afs client. I tried the -fakestat-all option and it did not work. Here is more background on the issue: I have an OpenAFS client that works after each reboot and then eventually hangs when the afs area is accessed. I am running openafs 1.4.9 that I compiled on the system (I have tried many versions of the openafs client with the same results). The OS is OpenSUSE 10.3. I have two other systems with the same OS that are working fine. I have searched /var/log/messages. I have checked the config files. User authentication works fine (even when the system is hanging). If I run the command 'ls -l /afs' that window is hung (or any other command that references afs). If an afs user logs in when the system is in a bad state the session immediately hangs because it can't cd to the afs home dir. I don't know what to do other than reboot. Can someone tell me what else to try to find out why this system is hanging? Thanks, Mark Henry Jeffrey Altman 04/29/2009 11:22 AM Please respond to jalt...@secure-endpoints.com To Mark Henry cc openafs-info@openafs.org Subject Re: [OpenAFS] /afs area is hanging Mark Henry wrote: > Thank you all for your responses. I am trying the -fakestat-all option on the > afsd daemon. We will see if it works. Everything works fine for awhile after > reboot and then it is just a matter of time before everything that touches afs > hangs. Hopefully this -fakestat-all option helps. > > Mark When the cache manager hangs, execute "cmdebug " and send the output. _ "This message and any attachments are solely for the intended recipient and may contain confidential or privileged information. If you are not the intended recipient, any disclosure, copying, use, or distribution of the information included in this message and any attachments is prohibited. If you have received this communication in error, please notify us by reply e-mail and immediately and permanently delete this message and any attachments. Thank you." _
Re: [OpenAFS] /afs area is hanging
Mark Henry wrote: > Thank you all for your responses. I am trying the -fakestat-all option on the > afsd daemon. We will see if it works. Everything works fine for awhile after > reboot and then it is just a matter of time before everything that touches afs > hangs. Hopefully this -fakestat-all option helps. > > Mark When the cache manager hangs, execute "cmdebug " and send the output. smime.p7s Description: S/MIME Cryptographic Signature
Re: [OpenAFS] /afs area is hanging
Thank you all for your responses. I am trying the -fakestat-all option on the afsd daemon. We will see if it works. Everything works fine for awhile after reboot and then it is just a matter of time before everything that touches afs hangs. Hopefully this -fakestat-all option helps. Mark _ "This message and any attachments are solely for the intended recipient and may contain confidential or privileged information. If you are not the intended recipient, any disclosure, copying, use, or distribution of the information included in this message and any attachments is prohibited. If you have received this communication in error, please notify us by reply e-mail and immediately and permanently delete this message and any attachments. Thank you." _ ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info
Re: FW: [OpenAFS] /afs area is hanging
I hate to state the obvious as I am sure you checked this already, but it really sounds like a firewall issue or something on the network is blocking traffic rather then the afs client itself. Sean On Wed, 29 Apr 2009, Wheeler, JF (Jonathan) wrote: > -Original Message- > > From: openafs-info-ad...@openafs.org On Behalf Of Mark Henry > > Sent: 28 April 2009 23:22 > > > > I have an OpenAFS client that works after each reboot and then > eventually > > hangs when the afs area is accessed. I am running openafs 1.4.9 that > I compiled > > on the system (I have tried many versions of the openafs client with > the same > > results). The OS is OpenSUSE 10.3. I have two other systems with the > same > > OS that are working fine. > > > > I have searched /var/log/messages. I have checked the config files. > User > > authentication works fine (even when the system is hanging). If I run > the > > command 'ls -l /afs' that window is hung. If an afs user logs in when > the > > system is in a bad state the session immediately hangs because it > can't cd > > to the afs home dir. I don't know what to do other than reboot. Can > someone > > tell me what else to try to find out why this system is hanging? > Thanks, > > If you really are using the command "ls -l /afs", I have always > understood that this is getting information about the whole of AFS > space, that is, the root directories for all AFS cells; naturally this > will take a long time. This may not solve your problem, but the command > as given is a bad (TM) idea. Of course, the command "ls -l > /afs/CELLNAME" where CELLNAME is the name of your cell is a much better > idea. > > Jonathan Wheeler > e-Science Centre > Rutherford Appleton Laboratory > (Cell rl.ac.uk) > -- Sean O'Malley, Information Technologist Michigan State University - ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info
Re: FW: [OpenAFS] /afs area is hanging
On Wed, 29 Apr 2009, Wheeler, JF (Jonathan) wrote: -Original Message- From: openafs-info-ad...@openafs.org On Behalf Of Mark Henry Sent: 28 April 2009 23:22 I have an OpenAFS client that works after each reboot and then eventually hangs when the afs area is accessed. I am running openafs 1.4.9 that I compiled on the system (I have tried many versions of the openafs client with the same results). The OS is OpenSUSE 10.3. I have two other systems with the same OS that are working fine. I have searched /var/log/messages. I have checked the config files. User authentication works fine (even when the system is hanging). If I run the command 'ls -l /afs' that window is hung. If an afs user logs in when the system is in a bad state the session immediately hangs because it can't cd to the afs home dir. I don't know what to do other than reboot. Can someone tell me what else to try to find out why this system is hanging? Thanks, If you really are using the command "ls -l /afs", I have always understood that this is getting information about the whole of AFS space, that is, the root directories for all AFS cells; naturally this will take a long time. This may not solve your problem, but the command as given is a bad (TM) idea. Of course, the command "ls -l /afs/CELLNAME" where CELLNAME is the name of your cell is a much better idea. It depends on your root.afs volume. With dynroot (as far as I understand), the involved effort is not great, and doing a plain listing of a static root.afs shouldn't take ages either. Most importantly, even if there is a longer batch of accesses that makes one process take forever, this should not freeze the entire machine's AFS connectivity, right? What happens if a user w/ AFS home logs on before ls - /afs? Does this not generate the bad state? Cheers - Felix ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info
FW: [OpenAFS] /afs area is hanging
-Original Message- > From: openafs-info-ad...@openafs.org On Behalf Of Mark Henry > Sent: 28 April 2009 23:22 > > I have an OpenAFS client that works after each reboot and then eventually > hangs when the afs area is accessed. I am running openafs 1.4.9 that I compiled > on the system (I have tried many versions of the openafs client with the same > results). The OS is OpenSUSE 10.3. I have two other systems with the same > OS that are working fine. > > I have searched /var/log/messages. I have checked the config files. User > authentication works fine (even when the system is hanging). If I run the > command 'ls -l /afs' that window is hung. If an afs user logs in when the > system is in a bad state the session immediately hangs because it can't cd > to the afs home dir. I don't know what to do other than reboot. Can someone > tell me what else to try to find out why this system is hanging? Thanks, If you really are using the command "ls -l /afs", I have always understood that this is getting information about the whole of AFS space, that is, the root directories for all AFS cells; naturally this will take a long time. This may not solve your problem, but the command as given is a bad (TM) idea. Of course, the command "ls -l /afs/CELLNAME" where CELLNAME is the name of your cell is a much better idea. Jonathan Wheeler e-Science Centre Rutherford Appleton Laboratory (Cell rl.ac.uk) -- Scanned by iCritical. ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info
Re: [OpenAFS] /afs area is hanging
You might also try the -fakestat on the afs startup daemon. On Tue, Apr 28, 2009 at 3:22 PM, Mark Henry wrote: > > I have an OpenAFS client that works after each reboot and then eventually > hangs > when the afs area is accessed. I am running openafs 1.4.9 that I compiled > on > the system (I have tried many versions of the openafs client with the same > results). The OS is OpenSUSE 10.3. I have two other systems with the same > OS > that are working fine. > > I have searched /var/log/messages. I have checked the config files. User > authentication works fine (even when the system is hanging). If I run the > command 'ls -l /afs' that window is hung. If an afs user logs in when the > system is in a bad state the session immediately hangs because it can't cd > to > the afs home dir. I don't know what to do other than reboot. Can someone > tell > me what else to try to find out why this system is hanging? Thanks, > > Mark > > > > _ > "This message and any attachments are solely for the intended recipient and > may contain confidential or privileged information. If you are not the > intended recipient, any disclosure, copying, use, or distribution of the > information included in this message and any attachments is prohibited. If > you have received this communication in error, please notify us by reply > e-mail and immediately and permanently delete this message and any > attachments. Thank you." > _ > ___ > OpenAFS-info mailing list > OpenAFS-info@openafs.org > https://lists.openafs.org/mailman/listinfo/openafs-info > -- David Bear College of Public Programs at ASU 602-464-0424
Re: [OpenAFS] /afs area is hanging
What do you get if you type "dmesg"? - Original Message - From: Mark Henry To: openafs-info@openafs.org Sent: 09-04-29 07:22:05 Subject: [OpenAFS] /afs area is hanging I have an OpenAFS client that works after each reboot and then eventually hangs when the afs area is accessed. I am running openafs 1.4.9 that I compiled on the system (I have tried many versions of the openafs client with the same results). The OS is OpenSUSE 10.3. I have two other systems with the same OS that are working fine. I have searched /var/log/messages. I have checked the config files. User authentication works fine (even when the system is hanging). If I run the command 'ls -l /afs' that window is hung. If an afs user logs in when the system is in a bad state the session immediately hangs because it can't cd to the afs home dir. I don't know what to do other than reboot. Can someone tell me what else to try to find out why this system is hanging? Thanks, Mark _ "This message and any attachments are solely for the intended recipient and may contain confidential or privileged information. If you are not the intended recipient, any disclosure, copying, use, or distribution of the information included in this message and any attachments is prohibited. If you have received this communication in error, please notify us by reply e-mail and immediately and permanently delete this message and any attachments. Thank you." _ ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info
[OpenAFS] /afs area is hanging
I have an OpenAFS client that works after each reboot and then eventually hangs when the afs area is accessed. I am running openafs 1.4.9 that I compiled on the system (I have tried many versions of the openafs client with the same results). The OS is OpenSUSE 10.3. I have two other systems with the same OS that are working fine. I have searched /var/log/messages. I have checked the config files. User authentication works fine (even when the system is hanging). If I run the command 'ls -l /afs' that window is hung. If an afs user logs in when the system is in a bad state the session immediately hangs because it can't cd to the afs home dir. I don't know what to do other than reboot. Can someone tell me what else to try to find out why this system is hanging? Thanks, Mark _ "This message and any attachments are solely for the intended recipient and may contain confidential or privileged information. If you are not the intended recipient, any disclosure, copying, use, or distribution of the information included in this message and any attachments is prohibited. If you have received this communication in error, please notify us by reply e-mail and immediately and permanently delete this message and any attachments. Thank you." _ ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info