vSAN RAID?

Similar Questions

• vSAN RAID controller compliance

  Hello

  I improved vSphere to 6U1/vSAN 6.1 last night and everything went very well - happy days. I noticed that the new plugin of the health has reported a compliance issue with my Raid Controller driver version. Either way, I'll have a latency intermittent host at this time and I think that this could be the culprit. I'm totally confused however as what needs to be installed. Before you deploy vSAN (about a month), I've updated through iDRAC all devices on my Dell R720xd 3 including RAID - H710Mini PERC controllers.

  vSAN THAT HCL needs:

  Firmware: 3.131.05 - 3415

  Pilot: megaraid_sas version 6.603.55.00.1vmw

  I need to reboot the servers to check the firmware version as iDRAC States 21.3.1 - 0004, which is apparently, according to this thread of dell, the version of the package, but assuming that the firmware version is correct, it leaves me with the driver update to version megaraid. I use a Dell ESXI ISO's so my question is, where can I get the megaraid driver and how do I install it in ESXI?

  Any help would be massively appreciated.

  That scrap. I've upated the "copy local HCL DB" and then ran a new test and everything is in line/green so everything seems fine!

• Cisco C240-M4SX / 12G SAS Embedded Raid Controller Queue depth Reporting resulting in ESXi 6.0

  All,

  We have been a customer of vSAN VMware 6.0 for the last 6 months of our environment entirely on servers of Cisco C240 M4SX with the Cisco 12 G SAS integrated Raid Controller.   Everything in the environment was working well until we started to bring in data warehouse loads in the environment and began to notice performance around latency of disk and most important issues still outstanding IO high.  After that ESXtop and ESXCFG examination we found that the length of the queue announced to the adapter ESXi has been only 234 however the VMware HCL Announces 895; 234 is below minimum spec of 256 to correctly implement vSAN.  We have worked diligently with VMware on it to try different versions certified and non certified driver async for this raid controller in addition to the most recent firmware for the raid controller.  Regardless of the change of the depth of the queue remained 234.  The presence of FBWC affect the depth of queue announced to the operating system?  VMware support has indicated that it is clearly a problem "hardware".  Any ideas as to what may be causing this?

  Our environment:

  C240-M4SX

  Integrated 12G SAS Raid Controller (operation JBOD / pass-through, not FBWC)

  UCSM 2.2 (6 c)

  Driver of VMware for controller: 6.606.06.00 - 1OEM.550.0.0.1331820.x86_64.vib

  VMware vSphere 6.0U1

  12 - 28-2015_5-51 - 31_pm.png

  

  12 - 28-2015_5-56 - 51_pm.png

  

  Greetings.

  Had the chance to spend some time in the laboratory and removed the cache module.

  After that, I now get "207", so I would say it's confirmed that 1,2,4 GB cache modules are used to increase the depth/length of the queue for these controllers.

  12gb_sas_qdepth_no_cache.jpg

  

  Thank you

  Kirk

• vSAN mini Dell H730 after update02 health

  Hello.

  I have 3 servers based on Dell H730mini. After the upgrade to esxi update02 vSAN was broke (because they were old drivers for the RAID).

  I have excluded all the vmware cluster servers, updated RAID drivers on compatible. After I created a new server cluster and it contains. Everything works.

  However, vSAN health module indicates the error type: "correct version of installed virtual san health invalid query. Relocation of vib module on each server correctly (not work), but in the cluster, all the information indicated the same version "in ESX 0.0.

  ESXi 6.0.0 38255889

  Vserver 6.0.03339083

  Any ideas?

  Hello, it seems to me that you must update vCenter.  Who is on the last version?  Thank you, Zach.

• Single controller RAID, LUNs, virtual disk groups question does?

  Hi VMware Experts.

  I have 5 servers (each at full load with local SSDS 8 x 1 TB).

  I'm curious:

  1. If I configure raid 5 (I'll have 6 TB per host), how many LUNS / data warehouses should create? Is it important here? In any case, we still have controller raid just unique, right?
  2. If I use VSANS, is there a difference in the answer above? A data store for each host is probably correct? Or because the VSAN configuration: raid0 /passthrough better to create several data warehouses/LUN?
  3. Insights into the performance here?

  Thank you!

  1. Yes, it is ok, because you have no other choice.

  2. you can't do several stores data vsan. A single vsan by cluster data store is possible.

  No matter how many ssd\hdd for each host that you have.

• VSAN 6.2 'method of fault tolerance' is not available

  Hello

  I installed 4 6, 0u2 ESXi hosts, each equipped with 3 flash devices and a VCVA 6.0U2

  I built the VSAN cluster without activating dedup and compression.

  But when I try to build a raid5 storage strategy, I can't find the parameter of the method of fault tolerance to allow coding of erasure with RAID5,

  I found the after fault tolerance method parameter enabled dedup and compression.

  Is this good?

  It's surprising. I have 60U2 basic vsan cluster where I am able to see 'Method of fault tolerance' the two option of raid 0 and raid 5/6 without activation of compression and deduplication.

  Can you confirm that you have all the flash disk group when you tried this option

  You can logout/login to the webclient service and check if you are able to see (there is a known, documented problem where some of the policies will not be visible unless you disconnect/connect back to the user interface)

  Was it installed or upgraded charges. If you upgrade, you did the upgrade of the disk version. I think that features will not appear if the upgrades of the disc are not (and automatically activating dedup/compression is an upgrade of the disk)

• Bandwidth VSAN not used with VM Home, but is used on hard disks

  Maybe it's just the way its supposed to be, I'm not sure.  I asked VMware support and they were very non-committal for trial on a response.

  I have a VM storage strategy that says: failure 1 host to tolerate, to allocate objects on 2 discs.

  vSphere to 6U1.  Freshly built and VMs and then migrate to it.

  When I look at a virtual machine, I see:

  VM House

  -Witness

  -Raid 1

  -component

  -component

  Hard drive

  -Witness

  -Raid 1

  -Raid 0

  -component

  -component

  -Raid 0

  -component

  -component

  Are only the VM host is supposed to follow a policy of storage defined?

  Or should I just use the built in Virtual Storage default SAN policy?

  Note: after the virtual machines have been put in this VSAN cluster (using my local storage strategy), I did go back and change the default virtual SAN storage strategy to have a bandwidth of 2 instead of 1.

  I don't understand why the VM host is not scratched as the hard drive is.

  Thank you!

  Paul

  This is an expected behavior for the VM host Paul Namespace.

  I wrote about it here - http://cormachogan.com/2014/03/12/vsan-part-18-vm-home-namespace-and-vm-storage-policies/

• RAID 1 @ Local controllers

  Is it possible to use RAID 1 controllers and detect and HDD and SSD in ESXi and use them in groups of disks VSAN?

  If you want to place a mirror RAID-1 around 2 SSD and present it to VSAN in a single volume for cache.

  While this can be done, there are a few considerations:

  1. cost. When you size it and pull your VSAN environment, you will have to double the cache requirements

  2. operational overhead. Volumes using RAID masks the fact that there is an SSD, you need to change the markup each volume as an SSD. In addition, when it comes to replacement of an SSD, you have many more steps to follow rather than disconnect the SSD drive and connect a new. Finally, the RAID volumes tend to hide his SMART stats, then you won't be able to query features like leveling of wear of the ESXi unless you have some special plugins RAID controller vendor. Finally, I do not think that you will be able to perform operations such as the flashing LEDs for easy identification, but again, this could be specific seller.

  These are the things you have to weigh up if you want to mirror SSD, and generally why we encourage it.

• Cluster VSAN configuration

  I would check a few questions about VSAN:

  1. How is starts up I can create about the cluster size?

  Scenario 1:

  Server 01:16 discs (2 SSD 1 TB, SATA 2 300 GB, SATA 1 to 12)

  Server 02:16 discs (2 1 TB SSD, SATA 1 to 14)

  In my example scenario 1, which is the ideal server for VSAN configuration? What I have to install ESXi on the 300 GB drive and the rest is used for the disk group as in Server 1? Or I can have a 2 server configuration since after installation, if applicable 14 discs I install ESXI, it will always be included when creating disk group?

  Scenario 2:

  If I apply VSAN with MRS, I created a data store separate space which is not included in the VSAN disk group or I can create within VSAN as a separate disk group (what I think is tedious because a diskgroup will include 1 SSD used for reserved space for a data store)?

  -Can I create starts with SSDS and SATA which are made up of different servers. As 3 SSD (1 each 1-3 Server) and 6 SATA drives (2 of each for 1-3 server) but they are grouped as 1 diskgroup?

  No, you can't.  The SSD and HDS support must be on the same host for a group of records.

  
  

  I see so if I installed the ESXi on this disc, it will be a full disc (I assume it will be the default value for the local data store seen first by the host) and not be included if you create groups of disks VSAN. Am I right? So it is best to have another drive to install ESXI.

  You are right.  A common practice is to install ESX on a SD card or USB stick, preferably in the opposite direction.  Attention to the good size.  Newspapers must not reside on the storage of vSAN and logging for vSAN will fill quickly a 8 GB or 16 GB SD/USB fast enough.

  -J' I always RAID on drives at the level of the material on the SSD SATA disks, and it is still necessary to right? If so, what VMware recommends RAID level? Can we raid SATA RAID 0, RAID 1 or RAID 6 disks? Or no RAID is necessary because VSAN will be pooling in a large data store (I guess it will be like a basic RAID software)?

  No, you do not have any RAID.  OK, you could do RAID 0 if you can not do passthrough.  vSAN needs to see the disks directly.

  
  

  -I guess I can create a data store local placeholder (another drive on the local server) to the replication vsphere.

  Yes, you could.

  
  

  Thank you, Zach.
  

• Drivers Intel P3700 for VSANS?

  I worked with our VMWare team to design a VSAN out and it included some Intel P3700 PCIe cards like the layer flash supported by about 1.2 TB of Seagate sas disks 10 k for storage.  Unfortunately, even of the P3700 disappeared not on the HCL, and I am currently "chasing my tail" with latency, and i/o pending.  I opened a SR, but recommended for the H730 drivers do not seem to do anything.  And I'm fairly certain that's not the issue (deactivation does not seem to be a problem).  I understand that the P3700 didn't HCL yet, but it is still ongoing and so many white papers were written with this card as part of the architecture.

  It's all running VMWare 6. 0 b (VMware ESXi 6.0.0 build 2494585 on hosts)

  Material:

  -From r630 Dell (30)

  -2 x Intel P3700 400 GB Flash cards (driver intel-nvme - 1.0e.1.1 - 1OEM.550.0.0.1391871.x86_64.vib)

  -10 x 1.2 TB SAS of Seagate 10 k disks

  -H730 Dell RAID controller in HBA mode with cache is disabled, etc. (drivers for this were something to worry about, but since I assume that the first writings go directly to PCIe cards, I don't think that's the question long - firmware 25.3.0.0016 pilot 6.606.12.00)

  -Intel X 710 with DAC 2 x 10 Gb connections

  -4 x Juniper QFX5100 10 GB switches (15 r630 by 2 switches, scaling up to 30 servers in 2 past later)

  
  

  SAN traffic exceeds 2 uplink (with uplink 1 standby), everything else is plus 1 uplink (with uplink 2 as before)-working on a plan to spend LACP, but it is still under construction

  
  

  The current workload is very low.  It is only about 6 Production VM running on that cluster to the repositories of the project and kicking the tires, with the idea that 500-1000 VM will be launched upward in the coming months.  For the majority of virtual machines, the performance is not essential, but some of the problems I see currently are a bit a showstopper.

  
  

  Problems:

  If I run "bonnie ++ u root" on a virtual machine single, I can see the latency time to go up to 65, 000ms (Yes, really 65 k ms) and the virtual machine is basically responding (100% iowait and very rarely is able to write i/o because of the huge latency).  The write buffer is never very full during this period (stuck at 30% and deactivation does not start even during the race.)  Similar issues occur ATTO Disk Benchmark running on a windows system with a disk queue high (4 seems to be well, 10 kills of the VM)

  
  

  I can get very high write speeds (500 - 800 MB/s or more), but as soon as the latency time leapt to a few 100 ms, it's all downhill.

  
  

  Even with a fairly simple recording VM with all our guests directed to these, I get occasional spikes of latency (1400ms +, with an average of 15ms, which seems high even).  This box just works a lot of entries to logstash and a hint of elasticsearch with occasional when bed kibana someone shows something.

  Y at - it no rider special I read somewhere for the P3700 Intel?  Another thing that I really should look into?  I'm tired of chasing my tail and want to start the actual load of the migration to this new cluster.  I tried RAID0 on a small cluster of 4 boxes, but that wasn't much better and is much more boring.

  I use the 1.6 and 2.0 TB P3700 cards in my VSAN without problem. I have a similar setup, except using r730xd, the same with the same firmware + drivers raid card, also on 6. 0b. When I originally put in place, I could see peaks of latency high 200-400 + ms, it was resolved by updating the drivers of NETWORK card for me. I use a NIC (Intel X 540 - AT2), different, but the firmware upgrade + the driver for my NIC brings my latency ~ 3ms avg and blip occasional peak at ~ 15ms. Probably worth to update any firmware on your X 710 and use the appropriate driver on VMWare HCL VMware Compatibility Guide: search for i/o device

  I've also seen really bad latency problems out of network configuration problem. In our case that we had a link to 1 GB failover in case our 10 GB failed, but instead things are started to be load-balanced and performance/latency was very poor until we noticed that the 1 GB has been fully utilized.

  Also try to use the Inbox NVME driver, I ran on it without problem before Intel released the driver 1.1 you use now. In my limited testing, Intel drivers are slightly faster performance but nothing majorly different.

• Laboratory VSAN with pizzaboxes with two bays?

  Hello

  I'm slightly older R300 servers Dell that I want to use a laboratory VSAN. Arrange it in two locations of drive and a RAID controller and I want to know if it is possible to use these for such a facility. I plan on keeping a 1 TB SAS drive in a slot and using a smaller SSD in the other, this configuration in a JBOD mode (in other words, doubt the controller supports true JBOD, so guess it'll be 'RAID-0 disk'). As I understand, ESXi must reside somewhere else, then maybe a USB pen drive?

  Is this possible?

  Thank you

  To?

  are there requirements on the size of the hdd/ssd?

  Yes, you need SSD of at least 10% of the size of the HARD disk drives in your disk group. Let's say that your have 1 TB HDD to add in a disk group, to support you need at least 100 GB SSD or more.

  How much space do I need for the root/ESXi?

  8 GB should be enough. If it is USB flash storage, smaller will do. because there will be zero persistent created by default.

• Some VMDK of getting scratched twice when the policy States 1 stripe. Disc format 6.0 VSAN 2.

  So I noticed this strange problem on my virtual SAN. 99% of my VMDK are properly implemented with 1 witness and a raid 1 array with 1 disk of two hosts. Like in this picture.

  

  Now, here is an another VM, with the same storage policy, creating a raid 0 of mirrors and to use more space than I care to provide to the virtual machine.

  

  I tried to reapply policies and nothing changes. It's not a huge deal, but I'm just curious to know what would cause this. Oh yes and the kicker is still its in line with the policy says VSAN.

  

  It's perfectly normal in a number of situations and is not a bug. For example, let's do all this time, a VMDK is larger than 255 GB. We are also when there is not enough disk space. See the series of Cormac Virtual SAN (VSAN) - CormacHogan.com blog and specifically part 23: http://cormachogan.com/2014/04/23/vsan-part-23-why-is-my-storage-object-striped/

  I also want to clarify that, contrary to what you said that a RAID-0 doesn't consume additional disk space. It is simply a way to split a mirror into smaller pieces. In short, the size changes however.

  Regarding compliance, if you find the definition, StripeWidth is a _minimum_ of batteries by mirror number to use. VSAN is free to use several axes. Note also in previous cases where we use RAID-0, we can place two strips on the same disc, as long that satisfy the minimum of StripeWidth pins of the requested user.

  Christian

• HomeLab vSAN woes

  Hi all

  I had a few problems of vSAN where readers will show degraded when in reality they are very well confirmed by LSI controller and the condition of the equipment. Readers (SSD) seem to become degraded when I/O strike records (that is, create a virtual machine). All the guests have 2 x the tag as a SSD HARD drive with one of these SSDS. I wrote a post on this subject in March here which has screenshots and more details. Life has been very busy (my 17 month around hunting ) finally have time to continue to troubleshoot. It's all in the vSAN HCL await real SSD drives.

  
  

  The hardware specifications are:

  Hosts:
  

  Servers Supermicro (X9SCM-F) x 2:

  • CPU: 2 x Intel Xeon E3-1230v2 'Ivy Bridge'
  • Motherboard: 2 x Supermicro X9SCM-F
  • Raid controller: 2 x internal LSI SATA/SAS 9211 - 8i
  • Memory: 2 x Kingston 32 GB Kit DDR3 1600 MHz PC3
  • Records: 2 x Lexar Echo ZX 16 GB
    • SSD: 4 x Sandisk Ultra II 240 GB
  • Network cards:
    • 2 x HP DDR Infiniband HCA 20Gbps adapter ports
    • 2 x HP NC360T PCI - e Gigabit Dual Port card
  • Power supply: 2 x 400W Seasonic 80 Plus Platinum Fanless ATX12V/EPS12V

  Supermicro X10SAE-o:

  • CPU: Intel Xeon E3-1231 "Haswell"
  • Motherboard: Supermicro X10SAE-O
  • Raid controller: LSI internal SATA/SAS 9211 - 8i
  • Memory: Kingston 32 GB Kit DDR3 1600 MHz PC3
  • Disks:
    • Lexar Echo ZX 16GB
    • SSD: 2 x Sandisk Ultra II 240 GB
  • Network cards:
    • HCA DDR Infiniband Dual Port Adapter 20Gbps HP
    • HP NC360T double Port PCI - e Gigabit card
  • Power supply:
    • SS-520FL2 520W ATX12V / EPS12V 80 MORE PLATINUM

  Any help would be greatly appreciated.

  Thank you

  Dave

  Apologize for the delay. During the last two day, I've updated my homelab to vSphere 6 (thanks to EVALExperience) and am happy to report vSAN works beautifully on the 20 GB infiniband network. I could create and clone virtual machines and SSDS are not going in an unhealthy situation. Thanks to everyone for their comments and suggestions!

• Experience the real world of vSAN to Production environments

  Hello

  I'm looking by putting forward a case to migrate an old HP EVA SAN company, to vSAN6 class. I'd really like some comments from those experienced with running the VSAN1 in their environments on its reliability and performance, and if they would consider worthy of a production environment VSAN6?

  Have you had a lot of pinch, or major problems with vSAN points?

  You give a target... we run in the order of 120 of average size Windows virtual machines in each of 2 data centers (which require 12 TB of storage in each ms) and would seek to 2 x 4 host VSANS loan node clusters.

  Would appreciate any input.

  Thank you

  We went with the config 'High' of Dell, with a few tweaks.  We have added 2 discs so that we had the max of 7 per disk group.  And we exchanged the Perc card a LSI 9207-8i.  Media LSI cross so we can hot swap drives when they fail.  PERC does not support RAID 0 to go down the server to share a drive.  As I said, in my rear view I would have 6 R720s instead of the R820s 3.

  Regarding the FTT, you can set per virtual machine so level 1 VMs get FTT = 2 and Test/Dev gets FTT = 1.

  Strange that you can not get them to the United Kingdom.  I guess you could just it build with your favorite retailer.  You do not have the convenience of 1 Reference, but you can get the right equipment.

  We have 2 1Gb dual port NIC for management and VM traffic goes to a stack of 3850 X Cisco switch.  We then 2 10 GB dual port NIC for vSAN goes to a stack of switch Dell 4032.  REALLY like the Dell switches, a lot of value for the price.

  Finished our about 60 k USD, but if I would have been to 6 knots, then each node would have about 35 k.  Your price sounds just for me.

  Thank you, Zach.

• vSAN on vSphere 5.5 - question on the compatibility of the motherboard

  Good afternoon, colleagues.

  There is a server with motherboards Intel S2600GZ, who is in the list of compatibility with VMware in the ' systems / servers.

  I will be sure in advance that it will work on vSAN if the external RAID should not be used (the disks connection will be directly on the motherboard)? Key word 'S2600GZ' or 'C602"section vSAN not looking for anything and he scares...

  Or, for example, take still - some Adaptek - external RAID, in the compatibility list is present, and in this vSAN manufacturer does not even. This means that any vSAN Adaptek will not work?

  The S2600GZ is on the HCL vSphere: link , you should be good there.  In regards to the vSAN, there are only three pieces of equipment in the HCL.  I/o Controller, hard drives and SSDS.  The team of the S2600GZ with the elements of vSAN HCL and it should be a supported configuration.  Note, support does not mean suitable for all workloads.  If it's a heavy load of work make sure you get the longer end I/O controller, HDD and SSD.  Stock of i/o Controller queue depth and support cross or not.  By the way you can share hot a disk failure.  Whether RAID 0 you have to down the host, replace the drive and re - create RAID 0.  For the SSD, see the class of performance AND endurance class.  Hard drives are pretty simple.  Sooner the better.  The more the better.  SAS over SATA.  Etc...  Hope this helps, Zach.