Interupting / breaking SMB transfers

[SMnasMAN]

Maybe this is a know issue, but i could not find anything on this:

There are quite a few operations, done via the web gui, that seem to cause destructive interruptions to (network) file transfers / sessions. (ie copying a file over SMB- being interrupted).

For example, im copying a few GB of data from a windows box to my FN server (via smb). On the FN box i go to create a new pool, unrelated to the pool im copying data to. At some point in the pool creation "in progress" the file copy from windows will be interrupted (network error message in windows, skip or try again. same result as if i stoped and started the SMB service ).

(ive seen the same issue occur when deleting a unrelated pool, or also with most other disk operations such as marking a disk offline, or adding a SLOG to a pool).

It also happens if you create a new SMB share. (this one makes more sense as i assume SMB service has to be restarted, but shouldn't there be a warning, to be sure there are no open files via SMB?)

Now that im aware of this, i know that, to be safe there are quite a few operations (or maintenance) that it seems you can only do if there are no open file operations (to be safe)

(anyone know if this also affect NFS or iSCSI? couldn't this cause corruption for cases like VM data-store shares - if the connection is randomly reset during a network IO operation )

in my 9th month of testing FN now, i have seen this occur quite a few times on entirely different test systems ive been running.

Is this a known issue? how is this acceptable without warnings or messages?

(im wondering if this cause issues on an very active FN box with many users accessing files, such as in an office).

Thanks for any input or confirmations.

 

[SMnasMAN]

I will update here (this reply) as i re-test additional functions, to see if they interrupt live SMB or NFS transfers.
(im testing with live transfers, like copying a file over smb or nfs- but i assume would have the same effect/interruption on an open file, such as if a user had an open excel file from a freenas SMB share)

- Tested adding a new dataset to a unrelated pool - (did NOT affect SMB transfer, nor NFS transfer)

 

[dlavigne]

Were you able to determine the reason for this? Which version of FreeNAS?

 

[SMnasMAN]

sorry, should have included that info:
Build FreeNAS-11.2-U5
Platform Intel(R) Xeon(R) CPU E5-2637 v2 @ 3.50GHz
Memory 262067MB

yes, Its still happening (maybe this is a known issue / just the way it is?). Also i know this same thing would happen on earlier versions of FN (on entirely different hardware) i ve been testing over the last 10 months or so.

as an example, today, i went to physically attach 3x new HDDs (supermicro disk bays), the disks showed up fine, with no interruption to NFS/ SMB transfers in progress. However as soon as i created a new pool with these 3x new disks, this occurred (see images):
(note the vmware CPU, is of a NFS backed VM running, so i can only assume that the dip was due to NFS dropping out for a bit as the timing matched up, however that one could be wrong/coinsidence. the smb is 100% when i clicked "create" on "volume manager" in fn web gui)

the FN box is a physical box (not a FN vm)

thanks

 

[seanm]

It also happens if you create a new SMB share. (this one makes more sense as i assume SMB service has to be restarted, but shouldn't there be a warning, to be sure there are no open files via SMB?)

I just tried to repro that, and could not. I started coping an 8 GB file. Then I created a dataset, made an SMB share for it, and changed dataset permissions. The copy did not stop, it completed successfully. I'm on 11.2-U5.

 

[SMnasMAN]

I just tried to repro that, and could not. I started coping an 8 GB file. Then I created a dataset, made an SMB share for it, and changed dataset permissions. The copy did not stop, it completed successfully. I'm on 11.2-U5.

thanks seanm, i had also confirmed that it does NOT happen with dataset creation, but my OP was about creating a entirely new, unrelated POOL (not adding a dataset) can you test pool creation? thanks

 

[SMnasMAN]

Well, to keep this updated: Today i added a 2nd vDEV to my large pool (pool name: he8x8TBz2 ).
and i saw NO interruption at all! (great)

I even had a non important SMB transfer going on, and that continued through with no issue.
Same for NFS shares on this pool ( NO interruption / problems - although i did suspend all VMs that use this pool, none of which have their os/boot disk on this pool).

not sure what to make of all this now, as i know on several prior occasions i saw interruptions as i detailed above.
will keep this thread updated. thanks

 

[SMnasMAN]

While i haven't tested this myself yet, it looks like the issue i was seeing has been fixed/addressed in v 11.2-u7:

NAS-102893 Avoid unnecessary Samba starts and stops when editing System Dataset configuration

Documentation Archive

Welcome to the Documentation Archive! All documentation provided here is end-of-life (EoL) and no longer receives any updates. TrueNAS SCALE TrueNAS SCALE 22.02 (Angelfish) Release notes for each minor release of 22.02 Angelfish are included in the archived docs. TrueNAS SCALE 22.02 (Angelfish)...

www.ixsystems.com

tks