4

To start thank you for your future help :)

  • I would like to know if it is possible to wrap my H.A as on my paint (excused me the quality of the drawing). My draw

  • I would also like to know how it goes with the ipv4 failover blocks in the case of an HA and a server that breaks down?

  • My last question, on my ISCSI if a hard drive crashes the data are lost if I understand correctly? so I would like to know if you have to do a raid or backups instead? without losing too much space ^^

Thank you in advance, I apologize for my English. cordially

HBruijn
  • 77,029
  • 24
  • 135
  • 201
Araxy
  • 51
  • 1
  • 1
    In general most RAID levels are designed to protect you against dataloss from a disk breaking but that is not the same as a back-up, which protects also against other forms of data loss, such as for instance an operator error. Typically you should have both. - As a minor issue but for future reference, please crop your images so it doesn't include so much white-space. - – HBruijn Sep 27 '17 at 07:57
  • Thank you for your answer, no problem for the pictures I would think – Araxy Sep 27 '17 at 08:18

3 Answers3

5

  1. Yes, you can do HA for the scenario you have described. To do so, you create an iSCSI target with multipathing and setup shared volume on top of it. Connect each compute host to target and check the failover. However, with regards to your diagram, the iSCSI storage is a single point of failure. If it fails, both compute hosts (server 1 & 2) lose the storage.

  2. To avoid data loss, you should have redundancy on each layer: disks, networking, nodes, rack, etc. The bulletproof setup for your case is a cluster of 2 computer and 2 storage node. Speaking about storage nodes, each one should have RAID installed and DAS configured in RAID 1/10/5/6 to obtain disk redundancy. RAID 10 is a choice of mine. The node redundancy can be implemented by shared-nothing storage like HPE StoreVirtual VSA and StarWind Virtual SAN is what I recommend here. The SDS does mirror storage: an each node contains the data. In case of failover partner keeps VMs running.

    From what I see, StarWind is certified with Xen and free. http://hcl.vmd.citrix.com/storage/712/StarWind_Software_Virtual_SAN

  3. I agree with other answers, RAID is not a backup. Additionally, to RAID and HA you should have considerations about RTO and RPO, backup solution implemented 3-2-1 backup strategy

Community
  • 1
Mr. Raspberry
  • 3,918
  • 13
  • 32
3

Q: My last question, on my ISCSI if a hard drive crashes the data are lost if I understand correctly? so I would like to know if you have to do a raid or backups instead? without losing too much space ^^

A1: You really need to rebuild what you have to have some replication between multiple independent "silos" of storage, for now you have SPOF you'd rather avoid.

A2: You need to have some backup (better two of them to follow what's called 3-2-1 backup plan) either way. RAID isn't a backup! HA replication isn't a backup replacement as well.

BaronSamedi1958
  • 13,676
  • 1
  • 21
  • 53
0

  • If I understand your drawing correctly, yes. On nic1 and nic2, if they are on 2 separate servers its perfect, as the nic slot0 and slot1 are used for the same thing on both server, Its the important point there. If both nic are on the same host you must create a bound.

  • Nothing happen in the stack, as the IP will come out of the other server if a server crash. Just be sure to have your switch port set to portfast or fastconnect. I had xenserver crash, and my users and I didnt even seen a cut.

  • Its like hbruijn told, a RAID is a capacity to run with a fail, its not a backup. Please plan a correct backup.

yagmoth555
  • 16,758
  • 4
  • 29
  • 50