Depending on the mechanism you’re using to copy files (if behind the scenes you’re using copy_file_range()) this may be a de-facto no-op since you’re just block cloning.
That’s a good point but it doesn’t explain the changes in the file system blocks pre and post script (see the sVDEV resource).
Before the change, most files were stored in 128k record size chunks, after, the vast majority were in the 1M range as one would expect after changing the dataset record size to 1M and running the script.
To me, that suggests the files were re-written?
As far as I saw, there is even a plugin that enhanced the capability to create ZFS datasets but, it was a community thing not an official contribution. Even though I kinda regret that when we run a docker compose and a folder needs to be created it will be created as a folder and not as a ZFS dataset I accept the decision 
Maybe in this sense: can directories be shown in the datasets section of the UI (maybe like a menu option)? I think for noobs like me, who did not realize that a docker compose created a directory and not a dataset is really helpful, and in general to have a quick overview of what is inside our pools
One option if you want a GUI for that is to install the filebrowser app, mount your pool under Additional Storage in the install screen, and use that to browse datasets and directories.
Ah, this is good to know! Thanks
on core, I used 2 commands the monitor performance…
zpool iostat -v 1
gstat -dp
Whats the equivalent for scale?
Bump… any idea what is triggering the critical boot pool alert in eel but not in dragonfish?
@Constantin is there a ticket filed for this? I was looking through to find one and couldn’t find it.
zpool command is the same.
for gstat try:
iostat -xm 2 10
zsh: command not found: zpool
is what I get from the shell logging in as truenas_admin
with the admin user you have to add sudo to the command so sudo zpool iostat -v 1
Just tested
zpool iostat -v 1 works fine with root, no need sudo here
This is clean install and only restore old settings
We recommend creating a non-root admin account in SCALE after uploading your CORE config file.
Yes, sorry. Forgot im in linux and not using root lol. that works now. we are just shocked at the increase in performance we are getting. the reads and writes are either double or quadrupaled. we are hammering it with a HUGE workload and it’s doing MUCH better than core. pretty wild. im impressed!
4 Likes
What kind of workload if you don’t mind me asking?
35 of our highest use terminal servers on a nfs store.
[quote=“Constantin, post:110, topic:12035”]
Bump… any idea what is triggering the critical boot pool alert in eel but not in dragonfish?
Good news, nightly from 9/20 should have a fix in for this. Will be in RC.1, will be testing today.
2 Likes
@kris my core installs are highly tweaked to get that speed too. this is just out of the box unmodified. makes me think I made a mistake on my tweaks lol
Wouldn’t be the first time we heard that, won’t be the last!
@kris for large systems with lots of memory and large l2arc is there some tweaking I could be doing?
i’m converting 2 other systems today. I did a quick test removing all my tunes, and I went from 6gb down to 2.5gb for speeds. I honestly think my bottle is my 12gb sas adapter now. thats insane.