[Solved] Checking Parity

[Mat1926]

I would like to initiate a parity check -last one like 2 months ago-, writing corrections is enabled by default, is this the correct way? Are there instances where it is advisable to disable that?

 

Thnx

[JorgeB]

Unless sync errors are expected, like after an unclean shutdown, you should always run a non correcting check.

[JonathanM]

When you run a non correcting check, any result other than zero sync errors requires further investigation. After you figure out why there are sync errors and correct the issue, then you must run a correcting check to bring the system back into sync. Some people feel that taking the time and effort to diagnose the issue isn't worth it, and recommend always doing a correcting check, but that risks data corruption if the source of the sync errors happens to be a data drive failure.

 

It's always better to do read only operations (non correcting check) until you are SURE you are changing the correct set of bits. A correcting check will only ever write to the parity drive, which may not be the correct thing to do in all circumstances.

[Mat1926]

@jonathanm @johnnie.black

 

Thnx 2 both of you for the information, I really needed that.

 

Can I ask you what will happen in case of power loss? Every time I reboot my system I manually start the array, so I assume the same is needed in case of power loss. In that case, the system will start parity checking automatically, correct? with or without corrections? what is the best way to handle the situation?

[JorgeB]

8 minutes ago, Mat1926 said:

Every time I reboot my system I manually start the array, so I assume the same is needed in case of power loss.

In case you don't know you can enable auto-start: Settings -> Disk settings

 

8 minutes ago, Mat1926 said:

the system will start parity checking automatically, correct? with or without corrections?

If array is set to auto-start it will start an automatic non correcting check, if auto-start is disable it will start a correcting check by default, you'd need to remove the "write corrections to parity" check mark before starting the array.

[JonathanM]

8 minutes ago, Mat1926 said:

Can I ask you what will happen in case of power loss?

Ideally your UPS will have enough power to allow a clean shutdown, and the array will be properly stopped.

 

Otherwise...

2 hours ago, johnnie.black said:

Unless sync errors are expected, like after an unclean shutdown, you should always run a non correcting check.

 

This is unclear in the context of what you are asking.

11 minutes ago, Mat1926 said:

Every time I reboot my system I manually start the array

Are you saying you make a habit of rebooting without stopping the array? If so, stop it. You stand a real risk of corrupting data if you reboot without cleanly unmounting things by stopping the array.

[Mat1926]

12 minutes ago, jonathanm said:

Are you saying you make a habit of rebooting without stopping the array? If so, stop it. You stand a real risk of corrupting data if you reboot without cleanly unmounting things by stopping the array.

 

isn't rebooting from the web cleanly unmounts everything and reboots the system?

 

*edit* Reboot will activate a clean system reset.

[JonathanM]

40 minutes ago, Mat1926 said:

 

isn't rebooting from the web cleanly unmounts everything and reboots the system?

 

*edit* Reboot will activate a clean system reset.

If you are pressing the button on the web GUI to reboot, that's fine. The word reboot isn't clear without context, as it just means to shut down and start back up. How you reboot makes all the difference.

[Mat1926]

6 minutes ago, jonathanm said:

If you are pressing the button on the web GUI to reboot, that's fine. The word reboot isn't clear without context, as it just means to shut down and start back up. How you reboot makes all the difference.

 

Let me thank you for your concern. When you use a modern OS you definitely don't press and hold the power button, the last time I did that I was using Dos 5.0/6.22...I always reboot/power off unRaid from the web gui using the proper buttons...

 

*edit*

You mentioned at the end sync errors should be 0...

 

 

[trurl]

4 minutes ago, Mat1926 said:

You mentioned at the end sync errors should be 0...

Zero is the only acceptable number of parity errors if you want to be sure that you can rebuild a data disk accurately. And of course, all other disks must also be good since all disks plus parity are required to rebuild a disk.

[Mat1926]

What will happen if that was not 0, what can we do?

[trurl]

4 minutes ago, Mat1926 said:

What will happen if that was not 0, what can we do?

Unless you have some specific reason to suspect a problem with a data disk, you will have to do a correcting parity check. After I correct parity, I always run another noncorrecting check since I want to be sure that my last parity check had zero sync errors and so there aren't any other problems.

[Mat1926]

2 minutes ago, trurl said:

Unless you have some specific reason to suspect a problem with a data disk

 

And that can be detected by checking the smart data?

[JonathanM]

1 minute ago, Mat1926 said:

 

And that can be detected by checking the smart data?

Hopefully. Also the system log. The diagnostics zip file is the best way to collect information about any error condition you are concerned about,  collected after the period where the problem occurred, and before rebooting if possible.

[trurl]

2 minutes ago, Mat1926 said:

 

And that can be detected by checking the smart data?

That is one thing that can make you suspect a data disk, and disks with SMART problems should show in the Dashboard with a warning.

 

But there can be other reasons to suspect a data disk, such as you just did a rebuild of a data disk. I always do a non-correcting parity check after a rebuild to make sure the rebuild went OK. If not I would consider doing the rebuild again rather than correcting parity.

 

Another reason you might suspect a data disk is if you were inside the case and might have accidentally bumped a cable connection, for example.

 

In the case of an unclean shutdown, it is generally assumed that parity errors are due to incomplete writes to parity caused by shutdown (power loss, etc) while writing to the array.

[Mat1926]

Just now, jonathanm said:

Hopefully. Also the system log. The diagnostics zip file is the best way to collect information about any error condition you are concerned about,  collected after the period where the problem occurred, and before rebooting if possible.

 

If these errors ever occurred, am I going to be notified by the usual email? I do receive occasional reports by email about my system condition/stats...

[Mat1926]

2 minutes ago, trurl said:

That is one thing that can make you suspect a data disk, and disks with SMART problems should show in the Dashboard with a warning.

 

But there can be other reasons to suspect a data disk, such as you just did a rebuild of a data disk. I always do a non-correcting parity check after a rebuild to make sure the rebuild went OK. If not I would consider doing the rebuild again rather than correcting parity.

 

Another reason you might suspect a data disk is if you were inside the case and might have accidentally bumped a cable connection, for example.

 

In the case of an unclean shutdown, it is generally assumed that parity errors are due to incomplete writes to parity caused by shutdown (power loss, etc) while writing to the array.

 

When the parity check is done w/o errors, I will update the disk settings so that the array starts automatically since it will auto check w/o corrections in case of unclean reboot...which I believe is the correct way to do after reading all your comments...

 

[trurl]

3 minutes ago, Mat1926 said:

 

If these errors ever occurred, am I going to be notified by the usual email? I do receive occasional reports by email about my system condition/stats...

If you have Notifications configured correctly, you will receive notification of SMART problems, and any actual read or write error due to other problems, but unRAID will often retry if there are problems and if the retry succeeds you won't get any email.