Linux, filesystem behavior mounted with errors=continue option?

Question

linux

Good afternoon
There is the following problem, there is a server with disks connected via a raid controller, it is used for recording video surveillance archives, the ext4 file system, occasionally some disks are weird and the system crashes, it is restored by rebooting the server. To at least not crash at all and be able to connect remotely, reboot mounted the partition with the continue option, but now I don’t know how to track the state of the partition in order to set up an alert or a script to reboot. Previously, in order not to reinvent the wheel, they would simply buy new wheels, but in the current situation it is very expensive ...

Answer 1

what raid? how many discs? what size disks?

occasionally some disks are weird and the system crashes

from bedblocks the system should not crash!
The video recording application may crash, but it can be restarted and it's faster than rebooting the entire system.
if it is a crash at the kernel level, then I strongly recommend removing the controller from the raid scheme or using it solely as a controller for connecting disks, and organize a software raid, since the problem may be in it and not in the disks
. Still, quite often, the cause of failures disks can be cables, I personally have already encountered this twice, and the disk behaved like the appearance of bad blocks, knocked heads, and smart had a bunch of dma errors.
In general, I advise you to first figure out what exactly is failing.

Answer 2

the raid controller is not needed for video surveillance from the word at all.
Especially if you have a problem with a hardware solution.
For video recording, there are more requirements for fault tolerance than for a caching system for the speed of data access, etc.
And do not forget that a hardware raid is the same software raid only in a piece of hardware with a separate stone, since the process itself requires a separate solution.
so this is a separate software complex, do
n't confuse it with a discrete video card anymore, it's a little different.
So the software is quite a normal solution, especially for your task.
If your task is NOT to master the budget.
The general io of the system in such a solution is not important at all, since it should only deal with video.

Answer 3

Hardware RAID is not always a good idea. I had iron RAID on both IDE and SCSI - they were very sensitive to voltage drops. Now network storages are often common and there are software raids, they are more reliable than hardware ones. And there are also special storages for video surveillance, there may be direct access to the disk, its own disk layout and encryption from above. In any case, "system restore" by restarting the raid is a very bad idea. We need means of monitoring the state of the raid and recovery. Most likely you'd be better off replacing the buggy piece of iron.