Copy speeds vary according to copy size

[Houseofdreams]

Hi

Noticed something strange. When I copy a single file (ex. around 1GB) the copy process to the truenas server goes flawless at max speed (+- 95% of 1Gbs)

But, when I try to copy a bunch of large files at once (let's say 5, all around 1GB in size) this happens

Constant dipping of the transfer speed, more times than there are files to copy, so it can't be some sort of delay between the files.

Truenas server:
- IBM M1015
- 4TB mirror WD drives
- 32GB ram

Any ideas what this could be?

 

[amichelf]

I am not sure if this has something to do with the way the Windows SMB client copies the data. If you are coping the data from aharddiskt it might be that your local disk might be the bottleneck. When you use robocopy with the /mt switch do you see different results?

Code:

robocopy src dest /mt:8

amichelf

 

[c77dk]

How far from eachother are the dips?

My guess is your disks might not be able to keep up and the dips are the TXGs being flushed to disk.

 

[Houseofdreams]

How far from eachother are the dips?

My guess is your disks might not be able to keep up and the dips are the TXGs being flushed to disk.

Hard to say, would need to time it :) I do have to say, I don't have a dedicated cache/zil/slog, as I think I don't really need this with my kind of data I think? (video files), so most files will be viewed once or twice, with a lot of time between.

Is there a way I can check these flushes in a log?

 

[HoneyBadger]

Hard to say, would need to time it :) I do have to say, I don't have a dedicated cache/zil/slog, as I think I don't really need this with my kind of data I think? (video files), so most files will be viewed once or twice, with a lot of time between.

Is there a way I can check these flushes in a log?

Try creating and running Adam Leventhal's dtrace scripts for checking dirty data amount and txg duration.

Show : dirty.d

Code:

txg-syncing
{
        this->dp = (dsl_pool_t *)arg0;
}

txg-syncing
/this->dp->dp_spa->spa_name == $$1/
{
        printf("%4dMB of %4dMB used", this->dp->dp_dirty_total / 1024 / 1024,
            `zfs_dirty_data_max / 1024 / 1024);
}

Show : duration.d

Code:

txg-syncing
/((dsl_pool_t *)arg0)->dp_spa->spa_name == $$1/
{
        start = timestamp;
}

txg-synced
/start && ((dsl_pool_t *)arg0)->dp_spa->spa_name == $$1/
{
        this->d = timestamp - start;
        printf("sync took %d.%02d seconds", this->d / 1000000000,
            this->d / 10000000 % 100);
}

Run them with dtrace -s scriptname.d and post some sample results during a copy showing the ramp from "fast" to "slow" to "stalling" - I bet your drives are having difficulty keeping up with a sustained 112MB/s.

How full is your pool, and what does a zpool list show for your free space fragmentation (FRAG%) column?

 

[anodos]

What version of FN is this?