VDI_IO_ERROR(Device I/O errors) when you run scheduled backup

shorian

I'll try a fresh install over the next couple of days and see if it reoccurs. Looking at the other boxes, I have the same error on one of the other hosts, but it's not across all of them despite identical installs and hardware.

Thanks for your efforts and help, shame there wasn't an easy answer but let's see if it reoccurs after a completely fresh install.

olivierlambert

Please keep us posted

shorian

Have just run a clean install - identical error in SMlog

Following was the last message in dmesg which seems counter to the SMlog error

EXT4-fs (dm-0): mounted filesystem with ordered data mode. Opts: (null)

Can send full SMlog and anything else you can think of - but it's either a hardware error (in which case why identical error accross two hosts?) or an incorrect setting in the install process. I'm at a loss

shorian

Have tried reformatting disks after changing the boot record - previously was GPT with ext4, have tried using MBR instead but no difference, even after another fresh install.

I'm going through a comparison with another machine that is identical kit but doesn't have those errors, but can't see any differences whatsoever

olivierlambert

Do you have a 4K sector disk?

shorian

The (large) SATA drives are, but the SSDs are not; problem is reported for both.

# parted /dev/sdb print

Model: AVAGO MR9363-4i (scsi)
Disk /dev/sdb: 6001GB
Sector size (logical/physical): 512B/4096B
Partition Table: gpt
Disk Flags:

Number  Start   End     Size    File system     Name  Flags
 5      1049kB  4296MB  4295MB  ext3
 2      4296MB  23.6GB  19.3GB
 1      23.6GB  43.0GB  19.3GB  ext3
 4      43.0GB  43.5GB  537MB   fat16                 boot
 6      43.5GB  44.6GB  1074MB  linux-swap(v1)
 3      44.6GB  6001GB  5956GB                        lvm

# parted /dev/sda print

Model: AVAGO MR9363-4i (scsi)
Disk /dev/sda: 500GB
Sector size (logical/physical): 512B/512B
Partition Table: gpt
Disk Flags:

Number  Start   End    Size   File system  Name              Flags
 1      1049kB  500GB  500GB               Linux filesystem

Also contrary to my thoughts above - the other machines do have the same errors in SMlog, only they've not been rebooted for many moons so it was in the archived off logs rather than current. So the problem is consistent across the entire estate.

olivierlambert

Before refocusing on the original issue, the SMlog error you have on reboot/boot, is it preventing you to use the SR?

shorian

No, everything seems to work just fine, except for the failure errors when running CR backups, and the messages in the logs when booting / attaching. The VMs themselves are performant, can be migrated on and off the host, and are handling production loads. If it wasn't for the CR errors, I would never have noticed them.

Meanwhile I've dug through the old Xen bug lists and found some similar SR issues but they seem to result in failure to mount but nothing that matches our specific situation.

olivierlambert

Hmm that's weird. I don't know what's causing this. Can you try with a local LVM SR and see if you have the exact same issue? (trying to see if it's related to the storage driver)

shorian

Fresh install, LVM on boot device, same error, created second disk as LVM through XO, same error, rebooted, same pair of errors.

Error appears to be agnostic of storage type.

Found a host with same drives, same CPU, same config, different RAID card (Adaptec instead of LSI), same error.

Error appears to be agnostic of storage manufacturer.

It seems to be the gc that causes it - elsewhere in the logs gc takes a lock, checks, decides there's nothing to cleanup, releases lock. Just when it boots, or when it is attached, its as though there is a slight delay where the storage reports as not being attached, fires the exception into SMlog, then gc tries again and connects just fine and gets the lock. We all continue as before, no other issues, no other errors, until either a reboot or a storage re-attach via CLI or XO. The EXCEPTION error does not occur during normal usage, nor during a backup.

Looking into xensource.log, we do get some errors as per below:

xapi: [ warn||1568 INET :::80|event.from D:87552104bbaf|xapi_message] get_since_for_events: no in_memory_cache!
xapi: [debug||1569 INET :::80|host.call_plugin R:91601df4a1d6|audit] Host.call_plugin host = 'c7634b1b-79b1-4eb2-840b-ecc401b7743d (xcp-ng-xen11)'; plugin = 'updater.py'; fn = 'check_update' args = [ 'hidden' ]
xapi: [ warn||1568 INET :::80|event.from D:87552104bbaf|xapi_message] get_since_for_events: no in_memory_cache!
xapi: [ warn||1570 INET :::80|event.from D:b4ea13ab245c|xapi_message] get_since_for_events: no in_memory_cache!
xapi: [ warn||1570 INET :::80|event.from D:b4ea13ab245c|xapi_message] get_since_for_events: no in_memory_cache!
xapi: [ warn||1571 INET :::80|event.from D:8b9b61c7b0b2|xapi_message] get_since_for_events: no in_memory_cache!
xapi: [error||1572 INET :::80||backtrace] host.get_by_uuid D:e4eb5b632522 failed with exception Db_exn.Read_missing_uuid("host", "", "df8fe71d-f748-4df4-bd53-97d8232249e4")
xapi: [error||1572 INET :::80||backtrace] Raised Db_exn.Read_missing_uuid("host", "", "df8fe71d-f748-4df4-bd53-97d8232249e4")
xapi: [error||1572 INET :::80||backtrace] 1/9 xapi Raised at file ocaml/database/db_cache_impl.ml, line 232
xapi: [error||1572 INET :::80||backtrace] 2/9 xapi Called from file ocaml/xapi/db_actions.ml, line 8179
xapi: [error||1572 INET :::80||backtrace] 3/9 xapi Called from file ocaml/xapi/rbac.ml, line 223
xapi: [error||1572 INET :::80||backtrace] 4/9 xapi Called from file ocaml/xapi/rbac.ml, line 231
xapi: [error||1572 INET :::80||backtrace] 5/9 xapi Called from file ocaml/xapi/server_helpers.ml, line 100
xapi: [error||1572 INET :::80||backtrace] 6/9 xapi Called from file ocaml/xapi/server_helpers.ml, line 121
xapi: [error||1572 INET :::80||backtrace] 7/9 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 24
xapi: [error||1572 INET :::80||backtrace] 8/9 xapi Called from file map.ml, line 135
xapi: [error||1572 INET :::80||backtrace] 9/9 xapi Called from file sexp_conv.ml, line 147
xapi: [error||1572 INET :::80||backtrace]
xapi: [ warn||1571 INET :::80|event.from D:8b9b61c7b0b2|xapi_message] get_since_for_events: no in_memory_cache!

However, given this only seems to occur at the point of boot or attaching the SR, I'm not convinced that this error is related to the VDI_IO_ERROR that we're seeing when running CR backups, whereupon xensource has the following:

xapi: [debug||257548 INET :::80|[XO] VDI Export R:2a08e12cc000|taskhelper] the status of R:2a08e12cc000 is failure; cannot set it to failure
xapi: [error||257548 INET :::80||backtrace] [XO] VDI Export R:2a08e12cc000 failed with exception Server_error(VDI_IO_ERROR, [ Device I/O errors ])
xapi: [error||257548 INET :::80||backtrace] Raised Server_error(VDI_IO_ERROR, [ Device I/O errors ])
xapi: [error||257548 INET :::80||backtrace] 1/13 xapi Raised at file ocaml/xapi/vhd_tool_wrapper.ml, line 75
xapi: [error||257548 INET :::80||backtrace] 2/13 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 24
xapi: [error||257548 INET :::80||backtrace] 3/13 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 35
xapi: [error||257548 INET :::80||backtrace] 4/13 xapi Called from file ocaml/xapi/export_raw_vdi.ml, line 49
xapi: [error||257548 INET :::80||backtrace] 5/13 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 24
xapi: [error||257548 INET :::80||backtrace] 6/13 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 35
xapi: [error||257548 INET :::80||backtrace] 7/13 xapi Called from file ocaml/xapi/export_raw_vdi.ml, line 86
xapi: [error||257548 INET :::80||backtrace] 8/13 xapi Called from file ocaml/xapi/export_raw_vdi.ml, line 110
xapi: [error||257548 INET :::80||backtrace] 9/13 xapi Called from file ocaml/xapi/server_helpers.ml, line 100
xapi: [error||257548 INET :::80||backtrace] 10/13 xapi Called from file ocaml/xapi/server_helpers.ml, line 121
xapi: [error||257548 INET :::80||backtrace] 11/13 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 24
xapi: [error||257548 INET :::80||backtrace] 12/13 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 35
xapi: [error||257548 INET :::80||backtrace] 13/13 xapi Called from file lib/backtrace.ml, line 177
xapi: [error||257548 INET :::80||backtrace]
xapi: [debug||258603 UNIX /var/lib/xcp/xapi||dummytaskhelper] task dispatch:session.logout D:f14c1d127f16 created by task D:4493412bc527
xapi: [ info||258603 UNIX /var/lib/xcp/xapi|session.logout D:4a1ab726bfb5|xapi_session] Session.destroy trackid=fc26bb0306ef13c29fa128ddb1338ca7
xapi: [error||257548 INET :::80||backtrace] VDI.export_raw_vdi D:4493412bc527 failed with exception Server_error(VDI_IO_ERROR, [ Device I/O errors ])
xapi: [error||257548 INET :::80||backtrace] Raised Server_error(VDI_IO_ERROR, [ Device I/O errors ])
xapi: [error||257548 INET :::80||backtrace] 1/13 xapi Raised at file lib/debug.ml, line 218
xapi: [error||257548 INET :::80||backtrace] 2/13 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 24
xapi: [error||257548 INET :::80||backtrace] 3/13 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 35
xapi: [error||257548 INET :::80||backtrace] 4/13 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 24
xapi: [error||257548 INET :::80||backtrace] 5/13 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 35
xapi: [error||257548 INET :::80||backtrace] 6/13 xapi Called from file ocaml/xapi/xapi_http.ml, line 193
xapi: [error||257548 INET :::80||backtrace] 7/13 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 24
xapi: [error||257548 INET :::80||backtrace] 8/13 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 35
xapi: [error||257548 INET :::80||backtrace] 9/13 xapi Called from file ocaml/xapi/server_helpers.ml, line 100
xapi: [error||257548 INET :::80||backtrace] 10/13 xapi Called from file ocaml/xapi/server_helpers.ml, line 121
xapi: [error||257548 INET :::80||backtrace] 11/13 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 24
xapi: [error||257548 INET :::80||backtrace] 12/13 xapi Called from file lib/xapi-stdext-pervasives/pervasiveext.ml, line 35
xapi: [error||257548 INET :::80||backtrace] 13/13 xapi Called from file lib/backtrace.ml, line 177
xapi: [error||257548 INET :::80||backtrace]
xapi: [error||257548 INET :::80||backtrace] VDI.export_raw_vdi D:2f552cd49c3a failed with exception Server_error(VDI_IO_ERROR, [ Device I/O errors ])
xapi: [error||257548 INET :::80||backtrace] Raised Server_error(VDI_IO_ERROR, [ Device I/O errors ])
xapi: [error||257548 INET :::80||backtrace] 1/1 xapi Raised at file (Thread 257548 has no backtrace table. Was with_backtraces called?, line 0
xapi: [error||257548 INET :::80||backtrace]
xapi: [error||257548 INET :::80||xapi_http] Unhandled Api_errors.Server_error(VDI_IO_ERROR, [ Device I/O errors ])

Two separate issues? The VDI_IO_ERROR is one that people would see through the GUI, so likely to be spotted, whereas the EXCEPTION in the SMlog is not going to be noticed unless you go digging - it seems to have no impact.

olivierlambert

I suppose there's no timing correlation between both? Indeed, the error in xensource seems more meaningful, because it happens during the backup process.

shorian

Nothing in SMlog of interest at the point of backup. We only started looking there when hunting for any SR related issues, but I’m inclined to view it as ancillary. The SR exception doesn’t seem to be impacting day to day operations, whereas the CR item is a big concern for us, and now we’ve dug into it further there doesn’t seem to be a correlation between the two.

So...I agree with your synopsis and propose putting the SMlog’s exception to one side and we concentrate on the original issue - with the xensource.log error from my previous post.

Tomorrow I’m going to see if I can find a way of simulating a brief access delay it and see if that recreates the problem - I’ve seen something vaguely similar in some old Xen 6 bug reports and I’ve run out of anything sensible to try. Strangest thing is that it’s consistent across all of the boxes, and survives a fresh installation - which reminds me, I’ll also try it from and to an lvm target.

olivierlambert

Thanks. That's a weird issue anyway, eager to know if you have the same behavior with local LVM.

shorian

Initial test - looks like it's working fine running CR from EXT4 to LVM. I'm now doing a deep soak to increase confidence in that statement, but so far so good.

Good - in that it narrows down where we look. Bad - in that I need the thin provisioning, but pleased we've got some progress. Good shout @olivierlambert

shorian

Soak test running between two local hosts, ext4 -> lvm has worked faultlessly. I've now run 10 complete backups, without a single issue. Previous doing this ext4 > ext4 would have thrown errors maybe 25-50% of the time.

However running between two remote hosts. ext4 -> lvm over a 1gbps connection has hit the VDI_IO_ERROR issue

I'll try the reverse, lvm -> ext4. lvm -> lvm might be a challenge but I'll try and get something organised; to rearrange the boxes a little.

shorian

@olivierlambert Update after 20+ iterations - writing to lvm (ext4 -> lvm) between remote hosts has only failed 1/20 times which for me is more than acceptable, compared with writing ext4 -> ext4 which fails 80%.

Given when it did fail on lvm it was the same error, I think it's less likely to be the driver as the root cause, but instead a timing or back pressure issue? All the more so given that the issue only arose when writing to a remote (outside the firewall) host, but writing to a host inside the firewall has not had a single issue.

Also I think it's interesting that the issue was (largely) resolved by making the change at the target end, not the source, despite there not being much in the way of error logs at the point of failure on the target.

So....where to look next? Do you any tools that can simulate the above so we can enable / disable and therefore isolate the problem?

(Happy to allow you in via XOA if it makes life easier)

olivierlambert

It's really hard to track those random issues.

Sounds like a kind of race condition happening on your hardware/setup.

shorian

@olivierlambert So chances of resolving are minimal to none. Ugh. I'll keep digging around, it's proving a right pain in the @£$%

olivierlambert

Well, if you spot something or able to reproduce it easily with some protocol other people could try and also reproduce, that would increase the chances a lot

shorian

Some more insight - for past couple of weeks I've been running a CR backup simultaneously to two SRs on different boxes. One consistently fails (remote to an identical XCP-ng box, but over the web on a 1gbps link) with the usual error, but the other succeeds (inside the firewall, 1gbps link, identical XCP-ng box), which would point to the problem not being at the source but the use of a wide area transit.

So...here is SMlog at the time of the failure on the target:

SM: [24325] vdi_detach {'sr_uuid': 'cf2dbaa3-21f3-903b-0fd1-fbe68539f897', 'subtask_of': 'DummyRef:|b73e41cc-a667-4ac0-b0f9-545dcd88b08a|VDI.detach', 'vdi_ref': 'OpaqueRef:3c20a324-2c22-4977-a63e-5baee37f9372', 'vdi_on_boot': 'persist', 'args': [], 'o_direct': False, 'vdi_location': '923acf0b-03bf-44ba-a846-f468141910f3', 'host_ref': 'OpaqueRef:d1eeb523-7938-4d80-a465-68ba2f149846', 'session_ref': 'OpaqueRef:a73dd5d5-2115-4bff-9d56-5cec26e267d8', 'device_config': {'device': '/dev/disk/by-id/scsi-3600605b010276170276bb1850660dd94-part3', 'SRmaster': 'true'}, 'command': 'vdi_detach', 'vdi_allow_caching': 'false', 'sr_ref': 'OpaqueRef:0cb18a22-646b-4dab-b08e-21ca36628f62', 'local_cache_sr': 'cf2dbaa3-21f3-903b-0fd1-fbe68539f897', 'vdi_uuid': '923acf0b-03bf-44ba-a846-f468141910f3'}
SM: [24325] lock: opening lock file /var/lock/sm/923acf0b-03bf-44ba-a846-f468141910f3/vdi
SM: [24325] lock: released /var/lock/sm/cf2dbaa3-21f3-903b-0fd1-fbe68539f897/sr
SM: [24380] lock: opening lock file /var/lock/sm/cf2dbaa3-21f3-903b-0fd1-fbe68539f897/sr
SM: [24380] lock: acquired /var/lock/sm/cf2dbaa3-21f3-903b-0fd1-fbe68539f897/sr
SM: [24380] ['/usr/sbin/td-util', 'query', 'vhd', '-vpfb', '/var/run/sr-mount/cf2dbaa3-21f3-903b-0fd1-fbe68539f897/923acf0b-03bf-44ba-a846-f468141910f3.vhd']
SM: [24380]   pread SUCCESS
SM: [24380] vdi_delete {'sr_uuid': 'cf2dbaa3-21f3-903b-0fd1-fbe68539f897', 'subtask_of': 'DummyRef:|6236329b-d253-4da6-bc9b-d5498fd02069|VDI.destroy', 'vdi_ref': 'OpaqueRef:3c20a324-2c22-4977-a63e-5baee37f9372', 'vdi_on_boot': 'persist', 'args': [], 'o_direct': False, 'vdi_location': '923acf0b-03bf-44ba-a846-f468141910f3', 'host_ref': 'OpaqueRef:d1eeb523-7938-4d80-a465-68ba2f149846', 'session_ref': 'OpaqueRef:b8fbaaeb-221f-4956-9088-f8565ba34d0a', 'device_config': {'device': '/dev/disk/by-id/scsi-3600605b010276170276bb1850660dd94-part3', 'SRmaster': 'true'}, 'command': 'vdi_delete', 'vdi_allow_caching': 'false', 'sr_ref': 'OpaqueRef:0cb18a22-646b-4dab-b08e-21ca36628f62', 'local_cache_sr': 'cf2dbaa3-21f3-903b-0fd1-fbe68539f897', 'vdi_uuid': '923acf0b-03bf-44ba-a846-f468141910f3'}
SM: [24380] lock: unlinking lock file /var/lock/sm/923acf0b-03bf-44ba-a846-f468141910f3/vdi
SM: [24380] lock: removing lock dir /var/lock/sm/923acf0b-03bf-44ba-a846-f468141910f3
SM: [24380] lock: opening lock file /var/lock/sm/cf2dbaa3-21f3-903b-0fd1-fbe68539f897/running
SM: [24380] lock: tried lock /var/lock/sm/cf2dbaa3-21f3-903b-0fd1-fbe68539f897/running, acquired: True (exists: True)
SM: [24380] lock: released /var/lock/sm/cf2dbaa3-21f3-903b-0fd1-fbe68539f897/running
SM: [24380] Kicking GC
SMGC: [24380] === SR cf2dbaa3-21f3-903b-0fd1-fbe68539f897: gc ===
SMGC: [24405] Will finish as PID [24406]
SMGC: [24380] New PID [24405]
SM: [24406] lock: opening lock file /var/lock/sm/cf2dbaa3-21f3-903b-0fd1-fbe68539f897/running
SM: [24406] lock: opening lock file /var/lock/sm/cf2dbaa3-21f3-903b-0fd1-fbe68539f897/gc_active
SM: [24406] lock: opening lock file /var/lock/sm/cf2dbaa3-21f3-903b-0fd1-fbe68539f897/sr
SM: [24380] lock: released /var/lock/sm/cf2dbaa3-21f3-903b-0fd1-fbe68539f897/sr
SMGC: [24406] Found 0 cache files
SM: [24406] lock: tried lock /var/lock/sm/cf2dbaa3-21f3-903b-0fd1-fbe68539f897/gc_active, acquired: True (exists: True)
SM: [24406] lock: tried lock /var/lock/sm/cf2dbaa3-21f3-903b-0fd1-fbe68539f897/sr, acquired: True (exists: True)
SM: [24406] ['/usr/bin/vhd-util', 'scan', '-f', '-m', '/var/run/sr-mount/cf2dbaa3-21f3-903b-0fd1-fbe68539f897/*.vhd']
SM: [24406]   pread SUCCESS
SMGC: [24406] SR cf2d ('Xen12 SATA') (18 VDIs in 16 VHD trees):
SMGC: [24406]         b1392757(5.000G/1.589G)
SMGC: [24406]         *d1789c22(150.000G/87.188G)
SMGC: [24406]             71e7c935(150.000G/313.500K)
SMGC: [24406]             aae21e12(150.000G/1.218G)
SMGC: [24406]         7b458e2e(10.000G/3.106G)
SMGC: [24406]         0fd9483c(25.000G/15.661G)
SMGC: [24406]         ae6739bf(10.000G/4.730G)
SMGC: [24406]         d25b8134(20.000G/3.759G)
SMGC: [24406]         265ed8d8(150.000G/16.609G)
SMGC: [24406]         32474df8(100.000G/98.434G)
SMGC: [24406]         5ae3d2e6(50.000G/46.990G)
SMGC: [24406]         7fc5f335(250.000G/16.885G)
SMGC: [24406]         ba7c92a3(25.000G/15.661G)
SMGC: [24406]         d6de6348(20.000G/9.978G)
SMGC: [24406]         24c6b7e8(20.000G/9.978G)
SMGC: [24406]         ab433fff(50.000G/49.595G)
SMGC: [24406]         6750bf45(10.000G/3.704G)
SMGC: [24406]         f9e3cd7e(10.000G/5.949G)
SMGC: [24406]
SM: [24406] lock: released /var/lock/sm/cf2dbaa3-21f3-903b-0fd1-fbe68539f897/sr
SMGC: [24406] Got sm-config for *d1789c22(150.000G/87.188G): {'vhd-blocks': 'eJzt1kESgiAUBmC5QTfgqB2Fq9QJOsKzCzTjqlooFakjkwjiQ5jm/xculHkf0MPUesil0zeNIAiCzOaa0T7rrhKBY4ld76pP3H7LLjr8Z2LI6Yv67cvEDLl9s/7UvjP9+nPx3v4L7PnoNk2x/7TKF13Z++8+mhzvhr7/FEOpSN+//rkQmy+ifK4M65fZfPEoYf35fHEqYv1NZj8wxOQOdYrZ/3/xxz5SrH7s5+Hd65vfX64qqszVPaVvNQopFbv/x9m7/ZSUDp6A5TcU+xZQv7fkdEZh/ups/wLZ//zZW5z//z+lT/4hpbz/1HCHFka7joc0V7sbl+qY58ryJ1FzVmCry7BhY3z9b8+iPVSsEb2/uU5mPzbw4cOHDx8+fPjw4cOHD38vv05U+QVg667e'}
SMGC: [24406] No work, exiting
SMGC: [24406] GC process exiting, no work left
SM: [24406] lock: released /var/lock/sm/cf2dbaa3-21f3-903b-0fd1-fbe68539f897/gc_active
SMGC: [24406] In cleanup
SMGC: [24406] SR cf2d ('Xen12 SATA') (18 VDIs in 16 VHD trees): no changes

The only thing I can see in there is the garbage collection.

Given that the issue seems to be something to do with the wide area network (same speed as inside the firewall, but obviously more hops), I wonder if this may be a similar problem as we're seeing in the S3 backups over the WAN? Could it be that some packets are being delayed in transit, or that there is an internal time limit on a backup chunk completing? Would also explain why backing up smaller VMs has a higher chance of success than the larger ones, since the % chance of a temporary timeout increases the longer a backup runs?

I've kept the set up as (finally) I now have 100% repeatability on the CR backup to the remote SR failure, so if someone wants to come in and have a look around it can be arranged. Others seem to have the same problem so hopefully this can help get to the root cause, as this the first time I've managed to get it to consistently fail every time. Appreciate that this doesn't sound like a success, but it beats an intermittent failure!