Benoire

Oh cool, that's great. Now that I'm running the system on vSphere, I can switch easily between the boot images to test boot and HDD detection. I also have a spare two drives in caddies that I can slot in when I can finally get to the install screen.

I thought I had uploaded the shots, but obviously not! Have attached two images showing this failure. Hmmm... can't upload them as board attachment quota is full...

Hmmm.. interesting, also getting this on vsphere too. I get an error 'fixing recursive fault but reboot is needed' after which the VM hangs. I currently utilise two HBAs; a Dell 6GBPS SAS2008 and a 'proper' LSI19211-8i. Both are flashed to IT mode using the latest firmware. Interestingly, if I boot with only the Dell HBA enabled it will load fine; utilising the LSI causes the above error and it hangs. The current DS boot image loads fine with both HBAs enabled and can load the drives fine.

Ok, so the image is having trouble with cpu hardlocks on my Xeon L5630 40w cpu in baremetal mode; the current DS boot image works fine but this one is refusing to load and gets hung up at various CPU hardlocks everytime. About to switch to vSphere 6 now so that should make it easier to test changes. Cheers, Chris

Hi Trantor, great. I'll give this a go over the weekend. Where would the boot logs be stored (or how can I extract the boot runtime data in to a log to post)? I've moving back in to vSphere soon to host Xpenology so will be even easier to test different boot images. Cheers, Chris

Command line will be the only way in which to do this as DSM doesn't have the ability to shrink an array. Personally, as MDADM can achieve this, its a bit of an oversight and would help will all the HDD size increases and changes necessary as time moves on... It stresses the drives out so I wonder if they didn't want to include another point of failure? Perhaps you should email Synology tech support and ask them if DSM will recognise the newly shrunk array if done via the CLI, if yes then you can do it that way.

I presume that using Diskstation on ESXi that I can insert a 'correct' synology MAC address for the network device in the virtual machine and not require the mac1, mac 2 etc in the config file and it will work fine?

Another thought, does the PAT file over-write the boot loaders synoinfo etc from the USB stick? If it does, then surely that would replace all the SAS stuff with default 3615xs stuff which doesn't support SAS? Doesn't get around the issues that its not detecting drives at the moment, but I wanted to ask.

@trantor (does this even work?!) So, put flash drive with SAS zimage in my supermicro chassis that has 3 LSI2008 cards (one true 9211-8i and two crossflashed dells to 9211-9i). Boots up fine, can find the IP addresses of the multiple NICs and then connect to the web gui, however it states that there is no hdd installed. I had one SATA drive installed in the machine and it doesn't get detected... So I would say that something is a miss here. Does the firmware of the LSI cards need to match the driver revision? My LSI cards are flashed to the latest P20 release from Avago but I wonder what Synology have done? Would they need to be the same version? Also looks like I'm still getting MPTSAS [FAILED] on boot, although now on the machine with 9211-9=8is it flashes past without halting... I'll drop back to my old usb but I'm not sure what else I can help with in terms of identifying failures or where its not going right. Chris

Ok so got an address, connects fine and brings up the install prompt, unfortunately the DELL SAS 6/IR installed, although and LSI 2008 chip, is not recognised... There goes my simple way of testing... Will have to drop the server down tomorrow, swap out usb's and use my spare 2TB drives to test and play with the new zimage, apologies!

Ok so I've got the boot image on the USB in the R410. It boots up, waits at MPTSAS for a while but I never see the Failed/OK and then gets to the final Diskstation login. Trying to find it on the network and its not there. LAN cable is connected and flashing but no IP address that I can connect to. Server worked on normal image, although after I've put my son to bed I'll try again with the normal boot image.

Sorry, is this image based on the 3615xs or a different model? I want to make sure I'm not using a 6.0 PAT file so can you confirm which PAT I should download. Thanks!

Great work Trantor. I've got a load of spare disks so will put them in my supermicro sas setup later, along with a the new bootloader (so I don't screw my current drives). Will be interested to see how it performs and whether it does what Synology have designed it to do in their systems. Its also good to see that the system can't detect the disks, suggests that it won't work without sas card. I've realised that I've got my spare Dell R410 sitting in my office being prepped for sale, it has proper 15K SAS drives installed so with 4 hotswap bays I should easily be able to test across SAS and SATA drives in multiple slots without any trouble! I'll do this tonight and provide any feedback!

Have a look on their forums as there are some threads that talk about it in more detail.

As far as I was aware, at present, Synology have built DSM using MDADM with LVM as the Raid / logical volume manager and then uses BTRFS as the file system over the top so it does not use the BTRFS inbuilt raid setup.

Thanks Trantor, wasn't sure if anyone would be that interested! I'm about to virtualise my xpenology build to give me 3 VMHosts to play with, once up and running I would be more than happy to test boot loaders. Is that patch for LSI cards in the bootloader? Is that to make it recognise LSI cards as sd(x) as otherwise its recognises as SAS and won't load as the synoconfig doesn't have the sas max enclosure commands in the config file? Can the boot loader inject a new synoconfig to override the one from the PAT file onces its installed? I presume not... I'm presuming we cannot patch a PAT file (I would imagine they have checksummed the file so after re-packing it, it would be a different size), so what would be required to support the loading of SAS drives and then the recognising of them? I'm presuming I can't just change the version file in the loader to boot up a rackstation PAT, and it would require the code to be compiled against that kernel?

image file, you can use wget from the command line to download directly to the box and then find it in root and then dd on to the usb stick from there... Any custom stuff e.g. mac and serial will need to be reapplied after updating as normal.

Yeah maybe a hash check on the length, easy way to check is for someone to create a new virtual machine with an older install and then edit the PAT and then try to install, if it works then we have a solution until such time that they incorporate verification.. If not, then spare HDDs or AllGamers method will by the way.

I found I didn't have to unmount the usb stick as its not used once booted. But the command you have listed below will work as I used that to do mine last week.

So the PAT file is a tar archive, so I guess if you extract HDA1 from within this archive and then change the synoinfo.conf and then repack as tar it should allow you to keep the 24 bay drive setup, worth a go in esxi?

Yes it would be. 7Zip can undo the PAT file but I'm not sure about rebuilding it. I've been looking through the PAT file for the rs18016xs+ and as that natively has SAS support and uses SAS HBAs and not SATA, that synoinfo.conf does not contain the portcfg line, instead it uses SAS_Enclosure_max which is set to 14... Would be interesting to see a boot loader that would allow the RS to load up as that might enumerate SAS backplanes properly & allow for more than 12 drives without editing...

Hi I've got a 16 bay SuperMicro chassis with a dual xeon motherboard (single Westmere L5630 installed presently), currently running a baremetal install of Xpenology. All is working fine, except that I cant get drives to be sorted according to their slot number, but that is an LSI hba well known issue. The purpose of this post is to ask what the benefits of virtualising DSM. I presently run 3 virtual machine hosts, 1 storage host (DSM) and a server essentials 2012 R2 box for TV stuff. Now, I can vrtualise this DSM machine easily and therefore add it to my host pool and load it with more memory, or I could combine a vmhost with this DSM box and remove the need for an additional machine. The answer to this will be decided by the various answers hopefully! So questions: 1) Is there any performance loss when utilising ESXi for DSM? 2) Anyone had any trouble with DSM not starting as expected when virtualised? Any issues that may crop up? 3) RDM vs passthrough? I've traditionally used passthrough but having looked at flexraid a while back which suggested RDM was the most flexible approach, what do people prefer and why? 4) If I use the array for VM storage, I assume that ESXi will reconnect to the array very quickly once its booted first? Now so more advanced case questions; My setup also contains vcenter and vsan so I cluster the hosts. 5) Any body run clusters with DRS/HA? Any issues there? 6) My primary VM storage array will be the vSAN once I get enough SSDs, presume no issues there too? My main question I guess is number 3, RDM vs Passthrough. ESXi should pick up the drives in the correct bays as I use SGPIO sideband connections from the backplane and using RDM will allow me to assign the drives in the right order to DSM, but I read that SMART is often not passed through to the VM and I would like to know more about that. Any thoughts, comments or suggestions on the above is very much appreciated. Cheers, Chris

Thanks, I suffered this on a new installation on my 16 bay unit until I realised. Does the boot loader contain any details around the number of bays etc. or is it simply a way to boot the kernel and prepare it for the new DSM update using the PAT file? Having looked at the PAT file, it obviously contains the synoconfig for the ds3615xs which is limited to 12 slots but I wonder if there is a way the boot loader can inject a modified one after unpacking? Probably not, but worth exploring?

Another thought, can we edit the boot image version to make it a rackstation rather than a diskstation as the rackstation is also bromolow?

Ok, so after more looking, I would be interested to know whether a boot image for an RS could be compiled? Does anybody know what is required to create a boot image for a rackstation? Do you need access to the an original rackstation or does the kernel and PAT file work? I wonder if getting a boot image based on the rackstations would give those with SAS drives full drive numbering based on the sas sequence rather than sd(x) sequencing which shifts for each drive depending on spin up. edit: you would need a boot image on further thought as I would imagine you'll need the drives setup as sas(x) during installation as changing the drive designation would probably put mdadm into a failure mode.