We Want to Hear Your Ideas

[danb35]

If iocage is the new direction in jails, as we've been told since ~9.10 that it is, then it needs to work. We now have four releases in a row with broken networking in iocage, all four of them broken in different ways. warden Just Worked(tm), whether that was because of warden natively, or wizardry in the GUI/middleware.

 

[Ericloewe]

I imagine that's part of the reason why FreeNAS going forward is going to use a branch of iocage, to better manage breaking changes.

 

[danb35]

Uh-oh, that's shades of The Release that Must Not Be Named... (specifically, its penchant for re-implementing everything from scratch)

 

[Ericloewe]

Uh-oh, that's shades of The Release that Must Not Be Named... (specifically, its penchant for re-implementing everything from scratch)

How do you figure? It's the same thing with the FreeBSD bits, they get forked over from Stable for each release (11.2 had its big merge a couple of weeks ago). There's no reimplementation here, or rather, there's no need for it.

 

[Borja Marcos]

Why not name your multipath drives with a meaningful name which describes their exact location in the system? I use gmultipath label -v J1S5 /dev/da0 /dev/da25 this tells me that this drive is located in JBOD1 Slot5. I have done this for years on many different systems including 90 bays and it works great. On systems with only one JBOD then I add a bit more detail to the name like part of the drives serial number and its enclosure ID for flashing lights to double check for example 5_78MH_2_4 tells me Slot5, serial, enclosure ID. On single pathed systems I do the same but with glabel.

That's a sort of Holy Grail but, unfortunately, due to hardware diversity it's not easy or even possible at all.

I remember that Sun systems had a nice way to identify drive bays. But, how do know which backplane slot is "da0"? LSI/Avago?/Broadcom HBAs, moreover, have the nasty habits of keeping a mapping between disk serial numbers and device identifiers.

The only reliable approaches I know, besides putting an actual, physical label, on each disk, is to use "sesutil map", "sesutil locate/fault", or even sas2ircu. And I would make sure that all of them work before relying on them. I have a Dell backplane here with some random shuffling of slot numbers. No idea why, it just happened at some time.

Welcome to peecee chaos at its worst.

 

[zha0hong]

A good percentage of these ideas people are posting are not FreeNAS ideas, they are FreeBSD ideas, or core ZFS ideas.

Ok, let's talk NAS ony. I think FreeNAS is getting worse for reliability when adding too much features. Our case is using FreeNAS Mini box as storage server for our remote offices over the world providing local storage and backups. replicate data between head office and remote sites. Joining windows domain was a pain during 6 years of supporting remote offices. sometimes works, sometimes dosen't. sometimes out of domain after reboot or upgrade. sometimes VMs or jails not working that I have to recreated. sometimes I have to reinstall FreeNAS or reset to factory and then I can join to the domain. that's why I have to be a FreeBSD expert that can modify your code or run shell commands to join to domain manually. even not counting how many times I have to manually run commands to replace a dead disk, while other products just need plug out a disk and then put in new one then rest will be done by system to recover a disk.

 

[Johnny Fartpants]

That's a sort of Holy Grail but, unfortunately, due to hardware diversity it's not easy or even possible at all.

I remember that Sun systems had a nice way to identify drive bays. But, how do know which backplane slot is "da0"? LSI/Avago?/Broadcom HBAs, moreover, have the nasty habits of keeping a mapping between disk serial numbers and device identifiers.

The only reliable approaches I know, besides putting an actual, physical label, on each disk, is to use "sesutil map", "sesutil locate/fault", or even sas2ircu. And I would make sure that all of them work before relying on them. I have a Dell backplane here with some random shuffling of slot numbers. No idea why, it just happened at some time.

Welcome to peecee chaos at its worst.

"But, how do know which backplane slot is "da0"?"

sas2ircu or sas3ircu will match the backplane slot with the drives serial number. Make a list of the which serial number matches which da number and start work naming your drives. After that it doesn't matter about the da numbers as your identifier is now the name of the drive itself.

In my case I find using the 90 bay Supermicro JBOD the system auto assigns da0 and da90 to drive slot one and then da1 and da91 to drive slot 2 etc so I don't generally go through the process of checking da numbers against serials as its not required however I always check the mapping afterwards before the system goes into production. On smaller systems its a quick job but on bigger systems it takes a while but just keep telling yourself you should only ever have to do this once.

I've done this for years and currently have thousands of drives setup this way on many different types of hardware with many replacements under my belt. Remember to do the same process when replacing drives. Single pathed systems are trickier as you need to name the drives and then make the pool on the command line and then export and import into the UI. Future disk replacements also need to be done on the command line otherwise you won't see the drive name after its been added to the pool.

Example of creating a pool on the command line using glabel names:

zpool create –f tank raidz2 label/1_50PV_2_0 label/2_ZHZV_2_1 label/3_52WV_2_2 label/4_TL7V_2_3

I normally create a dummy pool out of the drives before hand via the UI and then export it NOT MARK DISKS AS NEW as that way Freenas sorts all the partitioning for me hence why I need the force command above. Do the same for disk replacements ie. build a pool out of your new disk and then export and replace via the command line.

zpool replace –f tank label/1_50PV_2_0 label/1_16LE_2_0

(notice how only the middle part of the name has changed as that is part of the serial number, the slot is the same as is the enclsoure ID).

Your disk naming will not appear within the UI but it will on the command line with zpool status.

Once the pool is created via the command line set the following pararmeters.

From the command line set the zpool failmode to continue and autoexpand to on.

zpool set failmode=continue tank

zpool set autoexpand=on tank

From the webui set the top level dataset compression to lz4 as when you create the pool from the command line this does not happen by default.

Good luck and enjoy naming your drives.

 

[Johnny Fartpants]

That's a sort of Holy Grail

I call it the Rosetta Stone ;)

 

[acp]

I call it the Rosetta Stone ;)

Wednesday? :)


Sent from my Pixel 2 using Tapatalk

 

[majerus]

All, just wanted to let everyone know that we are still reviewing all the ideas. Thank you all for all the great ideas and feedback! I will be following up with this forum very soon!

Thank you,
Ben

Been a few months, just wondering what you guys are thinking regarding this thread.

 

[Johnny Fartpants]

Last idea from me. Auto shutdown of the server when you are on the brink of disaster. For example you have a 10 disk Z3 and you loose two drives the server auto-powers itself down. This could be a checkbox so that the people who don't won't it could simply uncheck. Granted you will need to power the server backup to address the issue but for those people who don't monitor their server or who are away for long periods of time then this could be a nice failsafe feature.

 

[dasti]

Ok, let's talk NAS ony. I think FreeNAS is getting worse for reliability when adding too much features. Our case is using FreeNAS Mini box as storage server for our remote offices over the world providing local storage and backups. replicate data between head office and remote sites. Joining windows domain was a pain during 6 years of supporting remote offices. sometimes works, sometimes dosen't. sometimes out of domain after reboot or upgrade. sometimes VMs or jails not working that I have to recreated. sometimes I have to reinstall FreeNAS or reset to factory and then I can join to the domain. that's why I have to be a FreeBSD expert that can modify your code or run shell commands to join to domain manually. even not counting how many times I have to manually run commands to replace a dead disk, while other products just need plug out a disk and then put in new one then rest will be done by system to recover a disk.

this is a very valuable experience, don't hesitate to send your suggestions !!

 

[dasti]

Built in OpenVPN in order to access FreeNAS and all jails.

and ipsec (i.e. openvpn doesn't pass the great firewall of china)

 

[MatthewSteinhoff]

We'd like the ability to have administrative users other than 'root' as well as change logging by username. If we could have two levels - admin and operator - that would even be better. (While admin would have all the abilities of root, operator may be read only.)

We have multiple system administrators. For all other systems, we're able to have a unique username for each admin. Most of our systems even have change tracking by username. Even those that don't have explicit change tracking at least provide a record of who logged into the system at what time.

(While it would be great if we could get all our admins to log every change themselves with a trackback to our internal ticketing system, that ain't happening. So, next best thing, I'd like to be able to figure out which admin made what changes. Then, when we do have a problem, we can look at the FreeNAS logs and see who was logged in around the time things went to hell.)

Cheers,
Matt

 

[ovizii]

and ipsec (i.e. openvpn doesn't pass the great firewall of china)

IMHO that would be best implemented as a plugin or docker as not everyone needs these features.

 

[mbalsam]

I would stop adding any features and put four people on your documentation full time. What should they do? Document every one of the recurring issues found in the forums. The developers should fix all of the reoccurring issues that are found.

- Active directory integration is a mess. Consolidate the existing samba documentation from Samba. NFS Permissions are horrible. I guess that's a unix thing, not really your issue.
- I searched for ZFS and got nothing. It's just an indexing issue, but really? ZFS primer is section 29? The link to your ZFS documentation is Oracle's broken link?
- Expand the (I) icons text in all areas.
- You need context-sensitive help, that appears outside of the main window.
- You need a list of all error messages and what needs to be done when each error is encountered.
- There is no log where you can track error messages seen in the UI in the file system logs.
- When do you look in /var/log/messages vs /var/log/debug vs /var/log/samba4/log.smbd etc.etc.

I don't really know the rest of the sections but I'm guessing that documentation is just as lacking..

I'm a developer so I can say this. Your letting developers design your product. Don't do it. You need a Product Manager who knows other products in the market to design your product. Dont use the "add features that are cool to the product" as your design. You need a solid, UI, Administration and authentication and error management.

You are so close to having an enterprise product, but skimping on documentation and assuming your SI partners will train the users is not a great strategy.

Sorry, it's tough love.

 

[RAurelian]

I would stop adding any features and put four people on your documentation full time. What should they do? Document every one of the recurring issues found in the forums. The developers should fix all of the reoccurring issues that are found.

- Active directory integration is a mess. Consolidate the existing samba documentation from Samba. NFS Permissions are horrible. I guess that's a unix thing, not really your issue.
- I searched for ZFS and got nothing. It's just an indexing issue, but really? ZFS primer is section 29? The link to your ZFS documentation is Oracle's broken link?
- Expand the (I) icons text in all areas.
- You need context-sensitive help, that appears outside of the main window.
- You need a list of all error messages and what needs to be done when each error is encountered.
- There is no log where you can track error messages seen in the UI in the file system logs.
- When do you look in /var/log/messages vs /var/log/debug vs /var/log/samba4/log.smbd etc.etc.

I don't really know the rest of the sections but I'm guessing that documentation is just as lacking..

I'm developers so I can say this. Your letting developers design your product. Don't do it. You need a Product Manager who knows other products in the market to design your product. don't use the "add features that are cool to the product" as your design. You need a solid, UI, Administration and authentication and error management.

You are so close to having an enterprise product, but skimping on documentation and assuming your SI partners will train the users is not a great strategy.

Sorry, it's tough love.

BEST COMMENT in this entire thread :) All the mentioned improvements would indeed be very very helpful!

 

[RegularJoe]

I agree Matt, we need better AAA on FreeNAS. I think I have started a thread here on the same topic as I have less skilled technicians that I do not want them logging in as root!

 

[dasti]

when setting up a new freenas, a "full automatic test" button with a report at the end : smartctl, memory test, iostat test, network test ..etc...
in my case it could be usefull before I ship a machine to a small office.

 

[orangehand]

I admit I am a total newbie to FreeNAS, but I have used and admin'ed Synology NAS's for a while.

What seems to be less than ideal:

Ease of AD integration - don't 90% of installs run AD? Or add an optional simple AD server a la Synology
A GUI method of syncing to the cloud, esp Backblaze
A better permissions system, to include ACL's (and I must admit to struggling with the perms setup generally, esp adding multiple groups to share)
Maybe add some sort of measurement of system impact when adding a VM, as I guess low cpu boxes could be easily crippled
Fully configurable email alerting if it's not there already - I need my customer's systems to alert me immediately if there's a work-stopping issue

Otherwise it seems Ace!!