代码改变世界

Linux服务器宕机案例第二则

2016-07-08 15:42  潇湘隐者  阅读(15649)  评论(3)  编辑  收藏  举报

    邮件告警发现海外工厂一Linux服务器连接不上，DPA（Database Performance Analyzer）系统也发现其出现问题，ping这台服务器发现网络不通，联系不到当地系统管理员，邮件咨询后，这个系统管理员也发现有问题，直接重启了，事后检查发现日志message里面，从10:10分开始出现下面错误信息（敏感信息处理了）

Jul  7 10:10:27 localhost kernel: ata2.00: qc timeout (cmd 0xa0)
Jul  7 10:11:00 localhost kernel: ata2.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x6 frozen
Jul  7 10:11:00 localhost kernel: sr 2:0:0:0: CDB: Test Unit Ready: 00 00 00 00 00 00
Jul  7 10:11:00 localhost kernel: ata2.00: cmd a0/00:00:00:00:00/00:00:00:00:00/a0 tag 0
Jul  7 10:11:00 localhost kernel:          res 51/20:03:00:00:00/00:00:00:00:00/a0 Emask 0x5 (timeout)
Jul  7 10:11:00 localhost kernel: ata2.00: status: { DRDY ERR }
Jul  7 10:11:00 localhost kernel: ata2: soft resetting link
Jul  7 10:11:00 localhost kernel: INFO: task hald-addon-stor:3728 blocked for more than 120 seconds.
Jul  7 10:11:00 localhost kernel: "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
Jul  7 10:11:00 localhost kernel: hald-addon-st D ffff8801989b6080     0  3728   3690 0x00000080
Jul  7 10:11:00 localhost kernel:  ffff880428dd76a8 0000000000000082 ffff880428dd7668 ffff88042c6300d0
Jul  7 10:11:00 localhost kernel:  ffff8804294ac0c0 ffff88040d730700 ffff8804294ac490 ffffffff812ef88e
Jul  7 10:11:00 localhost kernel:  ffff880428dd76a8 ffffffff81311442 ffff88042c630000 7fffffffffffffff
Jul  7 10:12:06 localhost kernel: Call Trace:
Jul  7 10:12:06 localhost kernel:  [<ffffffff812ef88e>] ? scsi_done+0x0/0x17
Jul  7 10:12:06 localhost kernel:  [<ffffffff81311442>] ? __ata_scsi_queuecmd+0x1a6/0x200
Jul  7 10:12:06 localhost kernel:  [<ffffffff814527e9>] schedule_timeout+0x36/0xe7
Jul  7 10:12:06 localhost kernel:  [<ffffffff812331b6>] ? kobject_put+0x47/0x4c
Jul  7 10:12:06 localhost kernel:  [<ffffffff8104367e>] ? need_resched+0x23/0x2d
Jul  7 10:12:06 localhost kernel:  [<ffffffff8145265b>] wait_for_common+0xb7/0x12c
Jul  7 10:12:06 localhost kernel:  [<ffffffff8104cf2f>] ? default_wake_function+0x0/0x19
Jul  7 10:12:06 localhost kernel:  [<ffffffff8121e449>] ? __generic_unplug_device+0x32/0x37
Jul  7 10:12:06 localhost kernel:  [<ffffffff81452773>] wait_for_completion+0x1d/0x1f
Jul  7 10:12:06 localhost kernel:  [<ffffffff8122540b>] blk_execute_rq+0xcb/0x104
Jul  7 10:12:06 localhost kernel:  [<ffffffff8121d88c>] ? freed_request+0x34/0x55
Jul  7 10:12:06 localhost kernel:  [<ffffffff812f60af>] scsi_execute+0xde/0x12e
Jul  7 10:12:06 localhost kernel:  [<ffffffff812f61cc>] scsi_execute_req+0xcd/0xf2
Jul  7 10:12:06 localhost kernel:  [<ffffffff81302a9a>] sr_test_unit_ready+0x5c/0xb5
Jul  7 10:12:06 localhost kernel:  [<ffffffff81302d0f>] sr_media_change+0x5d/0x28a
Jul  7 10:12:06 localhost kernel:  [<ffffffff810448c9>] ? task_rq_unlock+0x11/0x13
Jul  7 10:12:06 localhost kernel:  [<ffffffff8104cf08>] ? try_to_wake_up+0x3aa/0x3d1
Jul  7 10:12:06 localhost kernel:  [<ffffffff8131f73f>] media_changed+0x53/0x88
Jul  7 10:12:06 localhost kernel:  [<ffffffff8131f7a5>] cdrom_media_changed+0x31/0x37
Jul  7 10:12:06 localhost kernel:  [<ffffffff81303072>] sr_block_media_changed+0x19/0x1b
Jul  7 10:12:06 localhost kernel:  [<ffffffff8114407d>] check_disk_change+0x29/0x5b
Jul  7 10:12:06 localhost kernel:  [<ffffffff81322f9b>] cdrom_open+0x919/0x9b9
Jul  7 10:12:06 localhost kernel:  [<ffffffff810447f2>] ? __wake_up_sync_key+0x53/0x5f
Jul  7 10:12:06 localhost kernel:  [<ffffffff8142ebec>] ? unix_stream_recvmsg+0x40f/0x4d3
Jul  7 10:12:06 localhost kernel:  [<ffffffff813b63b5>] ? skb_release_data+0xaa/0xaf
Jul  7 10:12:06 localhost kernel:  [<ffffffff8142ebec>] ? unix_stream_recvmsg+0x40f/0x4d3
Jul  7 10:12:06 localhost kernel:  [<ffffffff8104367e>] ? need_resched+0x23/0x2d

clip_image001

 

系统版本为Oracle Linux Server release 5.7， VMware虚拟机，刚开始检查的时候，注意力集中在“ kernel: INFO: task hald-addon-stor:3728 blocked for more than 120 seconds”等相关信息上，关于hald-addon-store,这个是挂载媒体设备读写操作的进程，它是一个硬件守护进程，不停的检测有没有设备接入接出。

 

[root@localhost log]# grep 'Jul  7' messages | grep -E 'more than 120 seconds|hung_task_timeout_secs'
Jul  7 10:11:00 localhost kernel: INFO: task hald-addon-stor:3728 blocked for more than 120 seconds.
Jul  7 10:11:00 localhost kernel: "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
Jul  7 10:12:06 localhost kernel: INFO: task hald-addon-stor:3728 blocked for more than 120 seconds.
Jul  7 10:12:06 localhost kernel: "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
Jul  7 10:17:39 localhost kernel: INFO: task hald-addon-stor:3728 blocked for more than 120 seconds.
Jul  7 10:17:39 localhost kernel: "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[root@localhost log]#

 

而关于“task blocked for more than 120 seconds”，网上很多资料都千篇一律的说，修改参数 vm.dirty_ratio 和 vm.dirty_backgroud_ratio 可以避免这个问题。但是我对这个还是有点疑虑，理由如下一些：

 

1：这台服务器的vm.dirty_background_ratio 参数为10，vm.dirty_ratio为20。这个我查了一下，Redhat的官方文档介绍这是默认值。

[root@localhost log]# sysctl -a | grep 'vm.dirty'
vm.dirty_background_ratio = 10
vm.dirty_background_bytes = 0
vm.dirty_ratio = 20
vm.dirty_bytes = 0
vm.dirty_writeback_centisecs = 500
vm.dirty_expire_centisecs = 3000

 

2： 这台服务器已经运行好多年了，如果这个参数有问题，理应经常出现这个问题，但是到目前为止只出现了一次这个问题。

 

3： 我们还有多台服务器都是这个操作系统版本，内核参数同样是官方的默认值，也没有见到其出现过这样的问题。

当然也不排除某些操作触发了某些临界值导致了这个错误。那么我们还是继续分析一下错误信息吧，看能否挖掘出一些信息。

 

 

sar -r 查看发现这一段时间的内存使用情况，已使用内存的百分数（%memused）非常高，已使用交换空间的百分数（%swapused）比例也在个位数到40%直接，服务器在10:31重启，10:00 到10:30之间的数据没有，不清楚什么原因导致。从这些数据也可以看出内存资源非常紧张。

clip_image002

 

另外也用sar -u查看了CPU，CPU使用率不高。基本正常，当时忘记查看I/O和传送速率的统计信息（现在去查看，已经看不到昨天的数据了），另外在message日志里面看到有"ata2.00: status: { DRDY ERR }"， "ata2.00: qc timeout(cmd 0xa0)"; 之类的错误信息

clip_image003

 

关于这个错误，很多资料都说是硬盘故障或硬盘读写错误等等。也有说是内核bug的

clip_image004

 

当然我们系统的内核版本为2.6.32-200.13.1.el5uek ，这个https://lists.linuxfoundation.org/pipermail/bugme-new/2009-July/022389.html 链接也说了不止出现在2.6.31这个内核版本：

I have seen this messages not only on 2.6.31... also on many other 2.6 kernels.

但是自己的知识和经验也无法判断、定位问题，只能有个模糊的判断，要么是硬盘故障或读写错误，要么是内核错误。让那边系统管理员去检查存储设备是否有什么告警，暂时没有回馈，检查磁盘分区是否存在坏块，发现没有坏块情况

[root@localhost ~]# /sbin/badblocks -v /dev/sda
Checking blocks 0 to 104857600
Checking for bad blocks (read-only test): done                                
Pass completed, 0 bad blocks found.
[root@localhost ~]# /sbin/badblocks -v /dev/sdb
Checking blocks 0 to 157286400
Checking for bad blocks (read-only test): done                                
Pass completed, 0 bad blocks found.
[root@localhost ~]# /sbin/badblocks -v /dev/sdc
Checking blocks 0 to 41943040
Checking for bad blocks (read-only test): done                                
Pass completed, 0 bad blocks found.
[root@localhost ~]# /sbin/badblocks -v /dev/sdd
Checking blocks 0 to 104857600
Checking for bad blocks (read-only test): done                                
Pass completed, 0 bad blocks found.
[root@localhost ~]# /sbin/badblocks -v /dev/sde
Checking blocks 0 to 188743680
Checking for bad blocks (read-only test): done                                
Pass completed, 0 bad blocks found.

 

另外在message里面发现系统启动时，有大量BAR 13: can't allocate I/O resource [0x10000-0xffff]信息。

Jul  7 10:32:02 egvlnx03 kernel: ACPI: ACPI bus type pnp unregistered
Jul  7 10:32:02 egvlnx03 kernel: system 00:01: ioport range 0x1000-0x103f has been reserved
Jul  7 10:32:02 egvlnx03 kernel: system 00:01: ioport range 0x1040-0x104f has been reserved
Jul  7 10:32:02 egvlnx03 kernel: system 00:01: ioport range 0xcf0-0xcf1 has been reserved
Jul  7 10:32:02 egvlnx03 kernel: system 00:08: iomem range 0xfed00000-0xfed003ff has been reserved
Jul  7 10:32:02 egvlnx03 kernel: system 00:0d: ioport range 0xfce0-0xfcff has been reserved
Jul  7 10:32:02 egvlnx03 kernel: system 00:0d: iomem range 0xf0000000-0xf7ffffff has been reserved
Jul  7 10:32:02 egvlnx03 kernel: system 00:0d: iomem range 0xfe800000-0xfe9fffff has been reserved
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:15.3: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:15.4: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:15.5: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:15.6: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:15.7: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:16.3: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:16.4: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:16.5: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:16.6: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:16.7: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:17.3: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:17.4: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:17.5: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:17.6: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:17.7: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:18.2: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:18.3: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:18.4: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:18.5: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:18.6: BAR 13: can't allocate I/O resource [0x10000-0xffff]
Jul  7 10:32:02 egvlnx03 kernel: pci 0000:00:18.7: BAR 13: can't allocate I/O resource [0x10000-0xffff]

 

经过查证，有篇文章解释了这个消息出现的原因，并表示可以忽略这个问题。

  该消息表明内核无法分配预取内存插槽。

  预取内存插槽的编号是 13(PCI_BRIDGE_RESOURCES)。

  如果系统在 PCI 总线上有 13 个 PCI 桥设备，则会记录以上消息。

  虽然记录了该条消息，但是内核使用的是非预取窗口，而不是预取区域。 所以，可以忽略该问题。

 

内核参数我们暂时不打算修改，暂时只能先建议系统管理员增加内存资源，再观察系统具体运行情况。

 

参考资料：

 

http://ju.outofmemory.cn/entry/193909

https://www.blackmoreops.com/2014/09/22/linux-kernel-panic-issue-fix-hung_task_timeout_secs-blocked-120-seconds-problem/

http://linux.it.net.cn/e/bug/2016/0102/19226.html