STABLE train? Really?

[Ericloewe]

Yeah.. I like rolling versions in principle, but unfortunately it's usually more fun for developers to work on new functionality than on making stuff that "mostly works" even more stable, especially when the issues are rare or hard to debug. I don't blame them for this, just making an observation :p.My fear (which is too strong a word, but "concern" isn't quite right either) is that FreeNAS 10 will come out and FreeNAS 9.3-STABLE will basically be left in whatever state it's in at release.

Judging by the roadmap, 9.3 has a long time to become more stable than it is.

I do agree that a few major releases would be nice (kinda like what Microsoft is doing with several levels of updates, from cutting edge to security-only), however, 9.3 has been working well, with only minor inconveniences on occasion.

My suggestion is to choose a build you're comfortable with and stick with it for a while, skipping updates, if you don't want to stay with the latest STABLE releases.

 

[cyberjock]

Nope, 9.3 is going to be supported for a period of time after 10 comes out. My guess would be 6 months or so.

 

[Tywin]

Nope, 9.3 is going to be supported for a period of time after 10 comes out. My guess would be 6 months or so.

That's kind of my point.. 6 months isn't that long in the grand scheme of things, and "supported" can have a vast spectrum of meanings. For example, a longer support period, even without feature upgrades, would provide the impetus to make the product more stable (so that they don't have to spend as much effort supporting it during that time).

Ultimately it doesn't matter how long the support period is if the focus stays the same. The focus would have to shift to closing out existing bugs (not feature requests or enhancements, but actual bugs in features that are there now) in order for the 9.3-STABLE train to actually become more stable. I guess to be more pedantic about it, that is my concern, is that the focus won't change as 10 rolls around and everyone gets excited about it's new bells and whistles.

 

[DataKeeper]

As a Debian Linux user since version 0.93r5 (early 1995) I love their Stable, Testing & Unstable trees. Stable is just that.. Stable! I mean its rock stable.. period. It gets security updates or the rare critical patch and that's it. Testing is usually fairly stable but things can go wrong and updates can break things.. oh wait.. sounds like how FreeNAS 9.3-STABLE is.. or is it really FreeNAS 9.3-TESTING?

This is fairly disappointing to read about as someone new to the project and looking to dump a fair sum of cash into a system. If it's truly stable then whats being updated so often and why?

 

[cyberjock]

This is fairly disappointing to read about as someone new to the project and looking to dump a fair sum of cash into a system. If it's truly stable then whats being updated so often and why?

You don't have to ask the question. The update tells you the bug tickets and feature requests that were closed to that update. So you can read for yourself and see what's changed.

 

[DataKeeper]

Bugs and Security fixes I understand however feature requests have no place in a stable tree and when introduced without being fully tested breaks stable and gives you testing. If someone wants new features they should run FreeNAS from the Nightly tree where they should be introduced.

The 3 tree systems works great and takes very little extra resources to implement. Stable gets security patches and major bug fixes only keeping it stable. Add a new tree that that follows stable but adds feature requests as well as security patches and minor bug fixes. This is stated to be mostly stable but updates could break things temporarily. Continue with the nightly build tree, or unstable, working towards the next major release where you're on the bleeding edge of things.

This gives users (both free and paid) the ability to have a rock solid stable release that remains so between major releases and they don't need to worry about an update breaking things. I'm not saying it needs to be implemented today, this week or next month but it IS something iXsystems should take a look at and consider.

 

[mjws00]

I find this topic extremely interesting. In my mind, a filer is like a router, or a switch, or a dedicated server (think SQL, DNS, esxi). It is something you make BULLETPROOF and then you leave it alone. Critical security updates are acceptable, every other change adds risk to a stable and known good system.

I understand one can simply skip updates. But some type of vetted, truly stable entry point (even monthly or quarterly) would be smart imho. Frankly I don't want a random entry point onto a supposedly stable train to be one with a typo, or broken build system etc. The errors and quick and dirty patches are much too scary for production.

I'm still waiting for the dust to settle from the early adopters... but that day still hasn't arrived. I was one of the earliest on 9.3 testing... but still refuse to switch live servers, and you couldn't pay me to upgrade my pools at the moment.

I give iX all the props in the world, and I think this system allows incredible responsiveness and the ability to evolve quickly and address the user base. However, it leaves some of us folks that expect a BSD/ZFS combo to be rock solid in a lurch. This isn't like an app on my phone that auto updates and then breaks... It is dozens of TB of data at risk with significant penalties for recovery and down time.

I know I'm paranoid. That's what they pay me for. But I'm not even comfortable moving my personal, largely disposable, data at this point and that is almost tragic.

 

[Mlovelace]

In my mind, a filer is like a router, or a switch, or a dedicated server (think SQL, DNS, esxi). It is something you make BULLETPROOF and then you leave it alone. Critical security updates are acceptable, every other change adds risk to a stable and known good system.

I feel the same way; If I had to schedule an outage every time there was an update to 9.3, I would be pulling my hair out. These servers are, however, mostly backup targets for our production SANs, and are able to accommodate outages more frequently. That being said, a more "stable" train that only sees security updates and critical patches would be my preference.

I don't know how likely this is to change in the future though; as freeNAS is by-and-large the beta for trueNAS, and I think most freeNAS users would opt for a stable version. Which would mean less testing of feature requests/bugs fixes for trueNAS.

 

[zambanini]

well many would prefer to use truenas on supported hw. but buying hw from ix does not make any sense (at least in .de. at or. ch)

 

[DataKeeper]

....
I don't know how likely this is to change in the future though; as freeNAS is by-and-large the beta for trueNAS, and I think most freeNAS users would opt for a stable version. Which would mean less testing of feature requests/bugs fixes for trueNAS.

This is where the 3rd tree come in where any feature requests, new software, latest versions of plug-ins get added. Many folks would opt for this tree while leaving stable for businesses and home users who simply want a rock stable setup without the latest and greatest.

 

[jgreco]

This is where the 3rd tree come in where any feature requests, new software, latest versions of plug-ins get added. Many folks would opt for this tree while leaving stable for businesses and home users who simply want a rock stable setup without the latest and greatest.

Quite frankly, they haven't quite figured out how this sort of thing is best done just yet, and I also wonder if they have the developer resources to support multiple branches anyways.

I've always viewed FreeNAS as the bug-squashing version of TrueNAS. The bugs and some effort reporting them are part of the price you pay for the software, and if you look at it from that angle, it makes more sense. The guys in California are trying to make a living. They think they can actually do this by giving away FreeNAS while selling TrueNAS boxes. That means we get nice, good quality NASware (hella better than some supposedly "enterprise" products!) for free.

I decided long ago that this was still too good a deal and I decided that I'd try to spend some time providing support on these forums as part of my way of paying them back. My theory is that if I can make the software more successful on the FreeNAS side, FreeNAS is a gateway to upselling to TrueNAS, and then they're more likely to sell more TrueNAS, and be able to continue providing FreeNAS to me for free.

I fully understand that you'd like a totally stable FreeNAS, but I'd also encourage you to contemplate that it is still nice to pay the devs in some manner. Test a few updates. Make and support a plugin. Help provide support in the forums. Whatever.

 

[Tywin]

Quite frankly, they haven't quite figured out how this sort of thing is best done just yet, and I also wonder if they have the developer resources to support multiple branches anyways.

This, plus I'm not sure there are enough FreeNAS users that would ride the STAGING (or whatever you want to call it) train to really hammer on new fixes before the percolate to stable.

I've always viewed FreeNAS as the bug-squashing version of TrueNAS. The bugs and some effort reporting them are part of the price you pay for the software, and if you look at it from that angle, it makes more sense. The guys in California are trying to make a living. They think they can actually do this by giving away FreeNAS while selling TrueNAS boxes. That means we get nice, good quality NASware (hella better than some supposedly "enterprise" products!) for free.

I decided long ago that this was still too good a deal and I decided that I'd try to spend some time providing support on these forums as part of my way of paying them back. My theory is that if I can make the software more successful on the FreeNAS side, FreeNAS is a gateway to upselling to TrueNAS, and then they're more likely to sell more TrueNAS, and be able to continue providing FreeNAS to me for free.

Well, yes, but you have to treat both sides as a product. If you want lots of FreeNAS users to test your stuff for TrueNAS, you need to attract lots of FreeNAS users :) I'm not likely to run a FreeNAS box if I perceive it to be unstable and frequently breaking in minor ways with updates. However, if there were a rock-stable train that I could run, I would do that, but I would also be a lot more likely to have a VM somewhere running the STAGING train so that I can see what's coming down the pipe, meaning I would be testing things in the STAGING train in parallel with using the STABLE train for my actual data. Obviously not everyone would do that, but the general sentiment I've been seeing on the matter is that there are quite a few who would.

 

[jgreco]

I know. As I said, I don't think they've figured out how to do this well just yet. There'll be some people willing to work at each level of bleedingness, and some of us for whom stability is a large concern.

Since this is just a few months old at this point, patience is good. It can get a little messy along the way, but usually the end results here are better-than-average.