Shulou(Shulou.com)06/01 Report--

Some time ago, I received a response from my colleagues in Shenzhen that one of the oracle clusters was overloaded. In order to troubleshoot the problem, the system had been rebooted, but after the system restarted, six arbitration disks were lost.

In order to find out the reason, after logging in to the system, execute df-h, find that the / boot partition of the system has become mpathap1, enter the / dev/mapper directory, and find that there are three more mpathap1-3 files. The preliminary judgment is due to the fact that the system has its own multipath software. Check the initial configuration / etc/multipath.conf, which does not add the setting of shielding the local disk. After manually adding according to the document, because the local partition of the system is changed, the system needs to be rebooted, and it is found that the problem remains after the restart. After continuing to try the modification of / etc/fstab, the boot is directly bound to / dev/sda. As a result, the system restart is not normal and can only be modified by a single user. Try to uninstall the multipath software and restart, thought that the system will automatically return to the initial state, as a result, can not enter the system after reboot, need to enter the single-user mode, and reinstall the software. Go to / dev/ to see if sda1-3 finds that the system does not have this partition. It is finally confirmed that the system has been tampered with by multipath software after the system restart, and it is irreversible. If you solve this problem, you can only reinstall the system. So be sure to mask the local disk before configuring multipathing. And the probability of this problem is very small, because it has been rebooted many times before, so be careful when configuring it.

Through the execution of multipath-ll, it is found that the disk letters corresponding to the two wwid are out of order, and it really needs to be checked carefully later, which takes a long time, because there are 15 drive letters, and one pair is really troublesome. To avoid confusion of drive letters, / etc/multipath.conf has been modified to correspond to a fixed drive letter for each wwid.

Methods:

Multipath {

Wwid 36005076802810fb3000000000000007d

Alias mpathf

}

And then execute

Multipath-F clears the configuration.

Service multipathd restart restarts the service.

Multipath-v2 rescan.

Multipath-ll

At this point, the disk correspondence is solved, and the rest of the work is DBA.

Appendix adding method: execute / lib/udev/scsi_id-page=0x83-whitelisted-device=/dev/sda to view the corresponding wwid

Then add:

Blacklist {

Wwid 36234567890abcde01c6a312c0a867e33

}

In this way, the local disk is blocked instead of letting go of the default comments:

# blacklist {

# wwid 0x6234567890abcde01c6a312c0a867e33

# devnode "^ (ram | raw | loop | fd | md | dm- | sr | st) [0-9] *"

# devnode "^ hd [amurz]"

#}

This shields external devices such as optical drives. If the devices are directly shielded, it can also be written as devnode "^ sd [a Murz] *".

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.