May 8, 201511 yr I have several pro licences. At the moment I am attempting to install several small SSD's in my beta15 server. This server has now 1 parity and 22 data drives, I intent to fill the 23 slot with a data drive. When I now add a third SSD, then the previous SDZ get's kicked out and my server reports "configuration not valid". The SSD's are connected through MB SATA ports, so they get a sd* number before the data drives. It seems that it is not possible to have more than 26 drives! This is extremely limiting for me, because it would mean I can get either: - a cache drive - a cache drive but no VM's with their own dedicated drives. - 1 VM with a dedicated drive but no cache drive - JUST 1 VM with a dedicated drive - No SNAP drive if I want any of the above, and nothing from above if I want a SNAP drive. To use this 1 slot optimal, I would also need to buy a big SSD, 1 Tb or bigger, they are not cheap... With my WHS2011 VM, this would be the only extra drive I could use. So no cache or dockers on this server I'm not happy with this bug...
May 10, 201511 yr Buy a bigger drive and swap out some smaller ones. Or buy a RAID card (like the Areca 1230) and make RAID0 pairs out of the smaller disks.
May 10, 201511 yr Author Buy a bigger drive and swap out some smaller ones. Or buy a RAID card (like the Areca 1230) and make RAID0 pairs out of the smaller disks. You got some spare money?? At the moment the array consist of 1 x 6Tb parity drive, 4 x 6 Tb data drives and 18 x 4 Tb data drives. Thats 23 drives (96 Tb), the 24th drive is a 4Tb drive, ready to be inserted this month. I'll buy bigger drives, but they will be used in this server when needed (drives are all full), and the 4Tb drive will be re-used in my second server (I have 3 pro licenses). BTW: bigger drives are not the solution, I have an ARC-1200 card ready when I buy my first 2 8Tb drives, but I will be going from 23 data drives to 22 data drives (only 24 hotswap bays in this case). This would give only 1 extra drive slot for VM's, dockers (in total 2 drive slots). I also don't have the liberty to keep adding addon cards in this server since I'm using an A1SAM-2550F MB (1 x 8x slot for the M1015, 1 x 4x slot for the ARC1200 and the expander is located someware in the case (powered through a molex connector)). I would use ESXi (in hardware passthrough), but this MB also doesn't have vt-d. I bought this MB when nothing was known about a 26 drive limit, otherwise I would have bought the (more expensive) X9SCM-F-O so I could use ESXi.
May 10, 201511 yr You may have the largest unRaid server out there in terms of total capacity! The limitation has always been there. UnRaid was built to support three letter device designations - sda, sdb, ... When Linux gets to sdz, the next drive becomes sdaa - a four letter device name. UnRaid can't handle them. I've sometimes wondered if there was a way to get some of the disks to be recognized as "h" devices like the old IDE drives (hda, hdb, etc.) If there were, you would be able to overcome the 26 drives limit. Moving to 8T drives may be the answer, or start growing server #2.
May 10, 201511 yr You might also reconsider whether you want so many drives on a single parity system.
May 10, 201511 yr You might also reconsider whether you want so many drives on a single parity system. I know I get uneasy when I hit 10 drives on single parity despite having ability of going to 26.
May 10, 201511 yr Author The limitation has always been there. UnRaid was built to support three letter device designations - sda, sdb, ... When Linux gets to sdz, the next drive becomes sdaa - a four letter device name. UnRaid can't handle them. This is not true, read this post https://lime-technology.com/forum/index.php?topic=38189.msg353870#msg353870 , and then the thread from top till bottom.
May 10, 201511 yr Author You might also reconsider whether you want so many drives on a single parity system. I know I get uneasy when I hit 10 drives on single parity despite having ability of going to 26. Everything in life is a compromise. In this case, my uneasy feeling bites the dust against my wallet. This said, I will sacrifice a data drive when a second parity disk is available. But I won't double the amount of unraid servers I have (this is a (expensive) hobby, not a start as a datacenter)...
May 10, 201511 yr The limitation has always been there. UnRaid was built to support three letter device designations - sda, sdb, ... When Linux gets to sdz, the next drive becomes sdaa - a four letter device name. UnRaid can't handle them. This is not true, read this post https://lime-technology.com/forum/index.php?topic=38189.msg353870#msg353870 , and then the thread from top till bottom. When you bought a license, it is on our website that pro is for up to 25 devices. That is pretty clearly stated. Tom's reply on that thread says if it works, it works by accident for some, and it certainly isn't supported. In the future we may extend support beyond 25 drives, but we haven't reached that point yet. This is not a bug, it is a known limitation.
May 10, 201511 yr You might also reconsider whether you want so many drives on a single parity system. I know I get uneasy when I hit 10 drives on single parity despite having ability of going to 26. Everything in life is a compromise. In this case, my uneasy feeling bites the dust against my wallet. This said, I will sacrifice a data drive when a second parity disk is available. But I won't double the amount of unraid servers I have (this is a (expensive) hobby, not a start as a datacenter)... I had not seen that. It has been reported for as long as I have been here that 4 letter devices were not supported, but I had never tried it.
May 10, 201511 yr Author The limitation has always been there. UnRaid was built to support three letter device designations - sda, sdb, ... When Linux gets to sdz, the next drive becomes sdaa - a four letter device name. UnRaid can't handle them. This is not true, read this post https://lime-technology.com/forum/index.php?topic=38189.msg353870#msg353870 , and then the thread from top till bottom. When you bought a license, it is on our website that pro is for up to 25 devices. That is pretty clearly stated. Tom's reply on that thread says if it works, it works by accident for some, and it certainly isn't supported. In the future we may extend support beyond 25 drives, but we haven't reached that point yet. This is not a bug, it is a known limitation. I understand that we have 25 drives in the array. This is not a request for more drives in the array (either parity/data or cache(pool)). What I would like to do is use the new features that V6 brings us. Dockers, Vm's etc. At the moment I can do this in a very limited way. Just 1 drive for all my VM's and dockers. I was hoping, as an owner of several pro licenses, to use these new features in a somewhat more usefull (not limited to 1 drive per pro license) way. What does LT hope to accomplish with this limitation? If I knew this beforehand, I would have bought a vt-d capable MB and would thus circumvent this limitation. Using ESXi, I would have had as much extra drives/VM's as I have place in the case. I was hoping to use a somewhat cheaper, economical MB (less electricity bill), and to dump ESXi. After all, I would not need this with KVM and dockers integrated in V6! The current situation is for me a big disappointment. PS: I mailed a message to Tom about this limitation, I hope this will be solved in V6.
May 10, 201511 yr Wow. I'm amazed they even make motherboards that do not support VT-D. Oh, I see, it's a soldered on CPU that prevents VT-D.
May 10, 201511 yr What does LT hope to accomplish with this limitation? I know you are frustrated, so you probably haven't had time to think that through. Obviously, LimeTech had nothing to do with that limitation, as it's a decision that Linux made quite awhile ago, to use the 26 lower-case letters for drive identifiers. It was probably completely reasonable back then, because after all, "who could possibly want more than 20 drives in the same system"! It's the same as the Microsoft thing, way back when they designed the PC architecture. They weren't trying to accomplish anything by limiting the RAM, because after all, "who could possibly need more than 640K"! And it's the same thing as the Y2K problem. The year 2000 was so far off, it was unthinkable to us, and space was tight, so we used 2 digit years everywhere. We weren't trying to limit anyone, but needing more than 2 digits was beyond even planning for.
May 10, 201511 yr Author What does LT hope to accomplish with this limitation? I know you are frustrated, so you probably haven't had time to think that through. Obviously, LimeTech had nothing to do with that limitation, as it's a decision that Linux made quite awhile ago, to use the 26 lower-case letters for drive identifiers. It was probably completely reasonable back then, because after all, "who could possibly want more than 20 drives in the same system"! Please read very carefully this thread from top till bottom: https://lime-technology.com/forum/index.php?topic=38189 , the last message is very important... I'm not so much frustrated (at least not yet), more disappointed. Disappointed that I am not able to use the new V6 features like they are advertised. "Imagine having a workstation, gaming PC, media player, and a NAS, all running on the same hardware at the same time! Instead of buying multiple disparate systems for your various home computing needs, you can consolidate spending into a single, more powerful device. Seeing this first hand is extremely powerful and very exciting, as many folks have longed for this type of solution for quite some time…one Machine to Rule Them All!" "When we see some of the investments being made by our community members into much more capable hardware, we need to find better ways to take advantage of it!" I would like to use these new V6 features with my spare MB SATA ports: - 1 WHS2011 server VM with 2 dedicated 2Tb drives - 1 VM with a dedicated SSD as downloader - 1 SSD as docker drive/data drive for dockers I don't see any "fine print" that I can't do this with my PRO license, unless I give up some data drives... PS: This might come over as harsh and pushing, but I just want to get the facts straight...
May 10, 201511 yr The tell-tale screenshot is this one, showing device sdaa is assigned to disk 8. If this was possible in beta 12 (or whatever beta this was), i agree that it should not have stopped working in beta15. I am wondering if things happened in a peculiar order to enable this to happen. For example, disk8 may not have been sdaa when it was assigned to the array. It might have been sdr or something. And then after a reboot if it got assigned sdaa.
May 10, 201511 yr Author The tell-tale screenshot is this one, showing device sdaa is assigned to disk 8. If this was possible in beta 12 (or whatever beta this was), i agree that it should not have stopped working in beta15. I am wondering if things happened in a peculiar order to enable this to happen. For example, disk8 may not have been sdaa when it was assigned to the array. It might have been sdr or something. And then after a reboot if it got assigned sdaa. At the moment it is quite simple. Naming the drives is as follows (on my A1SAM-2550F): The UNRAID USB stick is sda Then comes all the MB sata ports Then the HBA sata ports are named If you have a full pro array with HBA cards (all 24 drives on the HBA+expander), then adding a drive on the MB sata ports gets first named sdb. The sdb on the HBA becomes sdc, and the sdy on the HBA becomes sdz. Add another drive on the MB sata ports, and this will be sdc. The sdc on the HBA becomes sdd and the sdz becomes ... (sdaa?) This results in a drive missing in the array, and you got a not working array. When I have time I'll try to rearrange the drives such that the empty slots are located on the expander (sdb-sde on the MB, sdf-sdy on the HBA). Maybe a sdaa drive can not be used in the array, but is still mountable as an "outside the array" drive. Still, this would not be an optimum solution, even if it works...
May 10, 201511 yr You might need to resort to customized persistent udev rules like the Areca RAID Controller users have to deal with. The modification would make drives show up as consistent lower enumerated devices based on drive serial-ids. At least that's the theory. I don't know if it's practical.
May 10, 201511 yr Please read very carefully this thread from top till bottom: https://lime-technology.com/forum/index.php?topic=38189 , the last message is very important... I'm not so much frustrated (at least not yet), more disappointed. Disappointed that I am not able to use the new V6 features like they are advertised. I don't see any "fine print" that I can't do this with my PRO license, unless I give up some data drives... I have fully read that at least twice before, and I read it all again, and it still hasn't changed, Tom's statement at the end is clear and correct. If somehow it appeared to work (and I'm not at all convinced that it did, completely safely), then it was an accident. I haven't seen the code, so could be jumping to assumptions too, but that variable looks like a 4 character buffer, storing a 3 letter drive symbol plus a terminating null, 4 characters in all. Until now, it has never needed more than a 4 character buffer, because drive symbols were always 3 letters. I believe that Nicktdot thinks a 4 letter drive symbol works, because Disk 8 in that pic had one, and it appeared to him that his server worked OK. He sees a working system, I see a buffer overrun! When I look at that pic, I'm scared for him, and I would shut the system down and remove a drive, because what I see is a byte of memory overwritten! A 5 character string was copied into a 4 character buffer, so a null probably overwrote the first byte of the following variable. That could cause any kind of bug or even a crash! I wouldn't trust that system with it. It does not surprise me AT ALL that a subsequent release did not work. And I'm sure I'm not the only veteran here that's shocked that it worked at all! I think Brian's idea is correct in how it appears to have been added. There hasn't been a need for "fine print". You have run into an unexpected limitation, unexpected to both you and us, because almost no one felt that anyone would need that many drive symbols. There is no direct limitation to using V6 features, and with a more normal number of drives, there would be no problem. And it's very clear from that thread you keep pointing to, that once Tom was made aware that a user or 2 was bumping into the drive symbol limit, he immediately put it on the schedule. And I suspect that the fact there is now another user hitting the limit, he may even raise the priority of implementing it. It won't be hard to recode, rather straightforward, but will take a little time. He will have to locate every variable that might be used to store drive symbols, and trace each one through all of the code, check its storage, its copying length, and its display. It's not just that its longer now, but that it's no longer a fixed length, could be 3 letters, could be 4 letters. Joe L will also have to check through the UnMENU code, and Brian will have to check the MyMain code, and possibly bonienl too, for any possible buffer overrun or variable display issue. At the moment it is quite simple. Naming the drives is as follows (on my A1SAM-2550F): The UNRAID USB stick is sda Then comes all the MB sata ports Then the HBA sata ports are named This is not correct, there is absolutely no guarantee what drive symbol you will get, even for the first, the flash drive. I took a look at 20 or 30 syslogs, and I have to confess I was quite surprised at how often the flash drive *was* sda! But just in that small set, I found 2 sdb's and an sdc, and I'm very sure that I have seen much higher letters. The Linux kernel appears to set up a number of threads for each controller and device, to initialize them and identify connected devices and capabilities, and assign them their slots and symbols. But it's a competitive world, and especially on more complex systems, there can be great variability in device response time, leading to very different assignments. The USB controllers do appear to be analyzed before the drive controllers, so it's not surprising that they often are first, but not always. Motherboard controllers too are quick to respond, and often but not always get the assignments early. But I have seen a number of cases where addon cards have beaten them. Edit: I reread the above, and my tone was not fully respectful, and I apologize for that. I sometimes think Tom gets unfairly criticized, for things out of his control. It's been my experience that he bends over backwards to help users, never places unnecessary limitations unless there are business reasons, and then they are both fair and clearly stated. When a new limitation appears, he fixes it, within his own time limitations. Give him time. He's always fair, and willing to help.
May 11, 201511 yr Author There hasn't been a need for "fine print". You have run into an unexpected limitation, unexpected to both you and us, because almost no one felt that anyone would need that many drive symbols. There is no direct limitation to using V6 features, and with a more normal number of drives, there would be no problem. I thought that 1 parity, 23 data drives and 1 cache drive is a normal number of drives for a pro license. Having to sacrifice the cache and several data drives for the V6 features is unexpected for me. It will also be costly, since I will have to buy a new MB, CPU, ram combo to be able to use ESXi. Because of this I'm disappointed in the current situation. Edit: I reread the above, and my tone was not fully respectful, and I apologize for that. I sometimes think Tom gets unfairly criticized, for things out of his control. It's been my experience that he bends over backwards to help users, never places unnecessary limitations unless there are business reasons, and then they are both fair and clearly stated. When a new limitation appears, he fixes it, within his own time limitations. Give him time. He's always fair, and willing to help. Writing rude and unrespectfull messages is also not my intention, they almost always don't help to solve the problem. But I do want to try to explain my point of view as good as I can. And doing so, I realize it can sometimes be seen as if I want to push a point of view, or seem to be a bit harsh in my comments. That's the nature of written communications (and I am not a native english speaker, so for me it's even more difficult). As I said in a previous message, I wrote an e-mail to Tom about this. I'm hoping this would be fixed in the 6.0beta releases (a man can hope, can't he ).
May 11, 201511 yr My understanding of this request is the following (but correct me if I am totally off): unRAID limits the number of devices in the array to 24 disks plus parity. This is a known design constraint and not being argued. The question is when more disks are put in the system for other purposes outside of the unRAID array, then these should not be counted. I believe v6 is the first version which really counts the total number of disks present in the system and stops when more than 26 are found.
May 20, 201511 yr An 'atttached device' limit was not intended for Pro keys. This has been fixed. In fact, you should be able to correctly use 4-letter device identifiers such as 'sdaa'. All places in the code have been corrected to allow for this now. But note: I have not tested these changes beyond 24 devices because I don't have enough hardware in house at the moment to do so You guys with huge arrays, let me know if seems to work... There is still a limitation of 25 combined array/cache slots. This restriction too will be lifted in a future release, but to do so now requires more changes and testing that we should do in -rc release.
May 20, 201511 yr An 'atttached device' limit was not intended for Pro keys. This has been fixed. In fact, you should be able to correctly use 4-letter device identifiers such as 'sdaa'. All places in the code have been corrected to allow for this now. But note: I have not tested these changes beyond 24 devices because I don't have enough hardware in house at the moment to do so You guys with huge arrays, let me know if seems to work... There is still a limitation of 25 combined array/cache slots. This restriction too will be lifted in a future release, but to do so now requires more changes and testing that we should do in -rc release. Very nice! Thank you very much!! 8)
May 21, 201511 yr Author An 'atttached device' limit was not intended for Pro keys. This has been fixed. In fact, you should be able to correctly use 4-letter device identifiers such as 'sdaa'. All places in the code have been corrected to allow for this now. But note: I have not tested these changes beyond 24 devices because I don't have enough hardware in house at the moment to do so You guys with huge arrays, let me know if seems to work... There is still a limitation of 25 combined array/cache slots. This restriction too will be lifted in a future release, but to do so now requires more changes and testing that we should do in -rc release. Thank you very much for looking into this. I'll be testing it as soon as I have some spare time...
Archived
This topic is now archived and is closed to further replies.