GUI very slow & and newbie questions

[Manticore]

Hey guys!

I am new to TrueNAS and new to the forums. I'm running into a few issues that I really need to resolve before making my move permanent.

First things first:
Server: HPe Gen9, Xeon E3-1225v5, 64GB ECC
Boot drive: Intel 80GB SSD (INTEL SSDSA2M080G2GC). Off-topic: 45,000 power-on hours, 100% health
ZFS pool: 4x HGST Ultrastar in 2x mirror vdev

After I start TrueNAS, I won't take long for the GUI to become very slow. Clicking something to pop up a menu can take up to a minute sometimes. Then when it finally opens, it can take another 10 seconds for the cursor to change to the finger-icon to indicate its ready for my click. Its rediculously slow :P. My last action led me to post my problem, because now it just hangs loading on 'edit' action on a container app. Refreshing the browser or reloading it doesn't help, it's stuck.

My searches has led me to slow response on the gui, but this always involves a usb drive (thumb or otherwise), and my boot disk is sata ssd (albeit kind of aged, like me ). I looked into it, I can test it, smartctl shows no problems, there's nothing in the logs / dmesg to indicate a problem (sata link errors, for example). That ssd is fine, and has always been fine with other Linux installations, so I don't think that's the problem.

What I do notice is a notification error sometimes, something with chart.release. Problem is, it shows in notifications but when I go to history / failed, its not there.

Another one is a python core dump. Also showed in notifications and popped up in my console. I copied that to a text editor so i saved it, but its also gone from history.

I noticed TrueNAS placed my swap space on a mdadm raid device spread over 4 spinning disks and created the pool using UUID's from a partition. I really don't want that, neither of that :P (it makes me itchy in places you don't want to know). If swap is really needed (I think swap is grossly overrated but let's not argue about that) I also really don't want to use partitions and UUID's for creating my pool.

I want to see this:

Code:

          mirror-0
            ata-HGST_HUH728080ALN600_ABCD1234               ONLINE       0     0     0
            ata-HGST_HUH728080ALN600_BCDE2345               ONLINE       0     0     0
          mirror-1
            ata-HGST_HUH728080ALN600_CDEF3456               ONLINE       0     0     0
            ata-HGST_HUH728080ALN600_DEFG4567               ONLINE       0     0     0

The reason is very simple: the disks are physically marked with that serial code (last 8 characters) and that makes it much easier to identify a failed drive. Having to write down UUID's and the corresponding serial code only makes things complicated and prone to errors. I know UUID's don't change between reboots, but technically it is possible to change them. Serials cannot change.

I can create a swap either in a swap file on the pool, or create a zvol for it.

So, in order to get what I want, can I delete that pool, remove the partitions and manually create my pool the way I want with a zvol for swap? Will TrueNAS pick that up? Will that be "supported"?

Last but not least, I cannot find any information about updates to TrueNAS and containerized applications. Will they retain their running configuration. I know a thing or two about docker and containers, persistent volumes and how I would save images with docker image save bla bla, but TrueNAS being an appliance with an exhaustive gui, I dont think thats the way. How should I handle a reinstall and what is the advised/proper way to restore containers with application settings?

Thanks everyone!

 

[KrisBee]

Have you read this? https://truecharts.org/manual/systemrequirements/

Swap on pool disks is the TrueNAs way, although you have the option to create swap on the boot drive at install time. I thought zvol for swap in linux has a bad idea, abandoned by Proxmox for instance a while ago. In anycase, it would be an unsupported hack.

 

[jgreco]

I noticed TrueNAS placed my swap space on a mdadm raid device spread over 4 spinning disks and created the pool using UUID's from a partition. I really don't want that, neither of that :P (it makes me itchy in places you don't want to know). If swap is really needed (I think swap is grossly overrated but let's not argue about that)

Your opinion is noted. However, running out of memory can kill the middleware or interfere with ZFS pool import operations, so it is not really optional. You can set the swap size to zero or place swap on the boot device if you wish, but it is mainly setting yourself up for potential problems down the road. If you want to do this, my suggestion is that you nevertheless reserve the space on the disks, because it can be super-difficult to rectify this down the road, like when the NAS is actually broken and unable to import its pool due to insufficient memory.

I know UUID's don't change between reboots, but technically it is possible to change them. Serials cannot change.

Sure they can. It just requires a slightly more specialized tool. This is a really weird argument. You can see the serial numbers in the disk list. This would make it impossible for you to dd a failing disk onto a new disk and have it just magically work like it should.

Will TrueNAS pick that up?

Maybe. TrueNAS will sometimes pick up broken or stupid things, but the general advice in such cases is to only rely on that long enough to export all your data off the pool and onto a backup, and then build a fresh pool using the GUI, and reimport the data, because this sets things up the way the appliance is expecting to see.

Will that be "supported"?

Airquotes-"supported"? Meaning can you come to the forums with problems caused by you designing your pool using a bespoke design that isn't what the system expects, and expect people to be able to help you, and for them to be happy when they find out that you've sabotaged your system? That's a resounding NO OF COURSE THAT'S NOT SUPPORTED.

If you mean might it work, well, yes, if it works, then it works, but it may only work until you need to do some operation where the middleware makes some underlying assumption as to how the pool is laid out or partitioning is designed, and then the middleware crashes or does the wrong thing. Building NASware is hard enough without users going offroad like this, and none of the developers over at iXsystems are hardening the system against all the possible ill-advised designs.

updates to TrueNAS and containerized applications.

At this point, TrueNAS SCALE is not a production version of TrueNAS, and you probably shouldn't rely on much of anything quite yet.

 

[Manticore]

Thanks for answering my questions, appreciate it.

For the record, I haven't changed anything from the default in my current installation, I'm exploring possibilities with regards to customising and setting it up more to my liking that's all. I have my answers, whether I like them or not. I didn't come here to argue and be some pedantic know-it-all, I apologise if I came across as such.

Having a default setup, is there anything I can look into with regards to the slow gui response and the python core dumps that seemed to have disappeared from history?

At this point, TrueNAS SCALE is not a production version of TrueNAS, and you probably shouldn't rely on much of anything quite yet.

I get that. Typically, a RC is close to being production-ready and I read RC2 is scheduled for december and a final version for februari, so I reckon it must be closer to ready than not at all.

If there are any ideas or suggestions on how to handle backups of containerised applications and their settings, I would really appreciate it.

Thanks again!

 

[jgreco]

I didn't come here to argue and be some pedantic know-it-all, I apologise if I came across as such.

No worries. Either you're open to feedback and can be taught, or you can't. We can't/won't change that. But it is generally a thing on these forums to make sure people make informed choices, so at a minimum, you walk away having been warned of the folly and what the safe recommendations are. Ideally, we want people to have the same enjoyable FreeNAS experience the rest of the forum members have had, which usually involves giving up some degree of freedom in the "UNIX cowboy" direction, but you get the reward of not having to design your own fileserver, Samba, NFS, monitoring, reporting, etc., all yourself. Only you can decide if that's worth it to you.

Having a default setup, is there anything I can look into with regards to the slow gui response and the python core dumps that seemed to have disappeared from history?

Is the slow response still happening? Is it slow if you access the system from the VGA console or via SSH? top report anything interesting? This didn't have any clear cause. If you're running containers or jails, these add complications to the system in that they can add extra pressure to a small system, but that's more like an 8GB or 16GB RAM system, where I'd expect swapping to be an issue.

I haven't been tracking the state of RC python core dumps, so, got nothin' for you there. I suspect that most of the likely problems have been discovered and reported, but you could take a look at current Jira tickets in the bug tracker if you like.

If there are any ideas or suggestions on how to handle backups of containerised applications and their settings, I would really appreciate it.

The floor is open to anyone who uses such things. We generally don't do that here.

 

[Manticore]

Is the slow response still happening? Is it slow if you access the system from the VGA console or via SSH? top report anything interesting? This didn't have any clear cause. If you're running containers or jails, these add complications to the system in that they can add extra pressure to a small system, but that's more like an 8GB or 16GB RAM system, where I'd expect swapping to be an issue.

Yes, it still happens now and then. It's not all the time but it seems to happen mostly when I work with an official or Truecharts app (I added that source for some apps). Currently, the only app installed is Syncthing from Truecharts. The other one I need is Duplicati. Maybe that's the one that was causing the problems, because it seems fast now but it got slow before I removed it. It could be coincidence too, as with that app it wasn't slow all the time (then it would be obvious). I would have to test more to be sure its one of those apps.

Terminal (ssh) is fast, no problems there. Top, htop, iostat don't show any cause. Load average in top is all well below 0.5. There is no iowait (0.3 once every 10 seconds or so), the system is empty and unused. k3s-server uses some cpu, goes up and down and averages around 10% with short peaks of 60% at most. Basically, nothing is happening on the system :).

To be clear, what I mean with 'containerised applications and their settings', I mean the Apps installed through the TrueNAS GUI, from the official or Truecharts app repositories. You're saying people tend not to worry about backing those up? I'm just worried that after a TrueNAS update, for whatever reason one of those apps won't retain its settings and I would loose their configuration. Maybe my concerns are unwarranted but I'd rather be safe than sorry. Just wanted to elaborate on that in case it wasn't clear what I meant with 'containerised applications'.

Cheers!

 

[jgreco]

You're saying people tend not to worry about backing those up? I'm just worried that after a TrueNAS update, for whatever reason one of those apps won't retain its settings and I would loose their configuration. Maybe my concerns are unwarranted but I'd rather be safe than sorry. Just wanted to elaborate on that in case it wasn't clear what I meant with 'containerised applications'.

I didn't say that. I don't really know what people are doing to back these up. The desire by hypervisor authors to dabble in storage, and the desire by NAS authors to dabble in virtualization/jails/containers, both of these are probably misadventures in unnecessary complexity, and there really aren't enough hours in the day to learn all the ins and outs of all these rapidly evolving and changing ecosystems, only to have that outdated in a year, or even next quarter.

Those of us who do these things at scale still tend to run pure NAS systems and separate hypervisors. When I spin up a web server today, I really just want to be able to do the same things I did a decade or two decades ago, and that's still quite viable. In the meantime, bringing up a FAMP stack on FreeNAS has gone through I think three different jail manager designs, possibly a hypervisor, and now on TrueNAS SCALE either a container or KVM, so that's like five or six different transitions in about as many years.

I am not saying this to be harshly critical of the FreeNAS/TrueNAS ecosystem, but rather to frame an answer to your question, which essentially revolves around the stability of the "app" ecosystem moving forward. Past history supports your idea that

those apps won't retain its settings and I would loose their configuration.

or that they'd even run, a year or two down the road. If you are installing non-trite things and investing anything more than a few minutes of time in setting them up, my own feeling is that you'd be better off making better hosting arrangements, such as a dedicated host for your containers.

I fully understand and appreciate the desire to commingle storage and compute, but the only thing that has actually shown itself to be vaguely stable over time is to run FreeNAS on ESXi. Even this has its caveats.

 

[Manticore]

Thanks, I'll see if some other answers pop up and then I'll weigh my options.