Storage Hardware for Hyperconverged oVirt: Gluster best practice
oVirt + Gluster (hyperconverged) RAID question: I have 3 nodes of SuperMicro hardware, each node has 1x SATADOM (boot drive for o/s install) and 6x 1TB SSD (to be used for Gluster). For the SSDs, is hardware or software RAID preferred or do I use an HBA? The RedHat docs seem to suggest hardware RAID, others on the forum say HBA or software RAID. What are other folks using?
Il 10 Gen 2018 20:40, <ovirt@fateknollogee.com> ha scritto: oVirt + Gluster (hyperconverged) RAID question: I have 3 nodes of SuperMicro hardware, each node has 1x SATADOM (boot drive for o/s install) and 6x 1TB SSD (to be used for Gluster). For the SSDs, is hardware or software RAID preferred or do I use an HBA? The RedHat docs seem to suggest hardware RAID, others on the forum say HBA or software RAID. What are other folks using? I'm using hardware raid. _______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Yaniv, You bring up a valid point. I asked about RAID since I was concerned about drive failures & performance. Since Gluster will handle data replication, using HBA seems like a better choice? >> From Yaniv: I think there are two interesting questions here: 1. Why would you want RAID? Your data is replicated already on other Gluster nodes. RAID writes will wear out the SSDs earlier and RAID 'wastes' additional space. The reason may be because you don't want the node to go down in case of a single disk malfunction. Fair - but is that the case? 2. I think HW RAID should be higher performing, but that's about it. Y. On 2018-01-10 12:05, Roberto Nunin wrote: > Il 10 Gen 2018 20:40, <ovirt@fateknollogee.com> ha scritto: > >> oVirt + Gluster (hyperconverged) RAID question: >> I have 3 nodes of SuperMicro hardware, each node has 1x SATADOM >> (boot drive for o/s install) and 6x 1TB SSD (to be used for >> Gluster). >> For the SSDs, is hardware or software RAID preferred or do I use an >> HBA? >> The RedHat docs seem to suggest hardware RAID, others on the forum >> say HBA or software RAID. >> >> What are other folks using? > > I'm using hardware raid. > >> _______________________________________________ >> Users mailing list >> Users@ovirt.org >> http://lists.ovirt.org/mailman/listinfo/users [1] > > > > Links: > ------ > [1] http://lists.ovirt.org/mailman/listinfo/users
On Fri, Jan 26, 2018 at 4:58 PM, <ovirt@fateknollogee.com> wrote: > Yaniv, > > You bring up a valid point. > > I asked about RAID since I was concerned about drive failures & > performance. > Since Gluster will handle data replication, using HBA seems like a better > choice? > Without local node replication (RAID), a failed disk == a failed node. This may or may not be acceptable. Y. > From Yaniv: >>> >> I think there are two interesting questions here: > 1. Why would you want RAID? Your data is replicated already on other > Gluster nodes. RAID writes will wear out the SSDs earlier and RAID 'wastes' > additional space. The reason may be because you don't want the node to go > down in case of a single disk malfunction. Fair - but is that the case? > 2. I think HW RAID should be higher performing, but that's about it. > Y. > > On 2018-01-10 12:05, Roberto Nunin wrote: > >> Il 10 Gen 2018 20:40, <ovirt@fateknollogee.com> ha scritto: >> >> oVirt + Gluster (hyperconverged) RAID question: >>> I have 3 nodes of SuperMicro hardware, each node has 1x SATADOM >>> (boot drive for o/s install) and 6x 1TB SSD (to be used for >>> Gluster). >>> For the SSDs, is hardware or software RAID preferred or do I use an >>> HBA? >>> The RedHat docs seem to suggest hardware RAID, others on the forum >>> say HBA or software RAID. >>> >>> What are other folks using? >>> >> >> I'm using hardware raid. >> >> _______________________________________________ >>> Users mailing list >>> Users@ovirt.org >>> http://lists.ovirt.org/mailman/listinfo/users [1] >>> >> >> >> >> Links: >> ------ >> [1] http://lists.ovirt.org/mailman/listinfo/users >> >
If a disk fails (aka node fails) assuming no RAID, once the disk gets replaced, the node will rebuild? Is that what will happen or am I over simplifying? On 2018-01-26 09:02, Yaniv Kaul wrote: > On Fri, Jan 26, 2018 at 4:58 PM, <ovirt@fateknollogee.com> wrote: > >> Yaniv, >> >> You bring up a valid point. >> >> I asked about RAID since I was concerned about drive failures & >> performance. >> Since Gluster will handle data replication, using HBA seems like a >> better choice? > > Without local node replication (RAID), a failed disk == a failed node. > This may or may not be acceptable. > Y. > >> From Yaniv: > I think there are two interesting questions here: > 1. Why would you want RAID? Your data is replicated already on other > Gluster nodes. RAID writes will wear out the SSDs earlier and RAID > 'wastes' > additional space. The reason may be because you don't want the node to > go > down in case of a single disk malfunction. Fair - but is that the > case? > 2. I think HW RAID should be higher performing, but that's about it. > Y. > > On 2018-01-10 12:05, Roberto Nunin wrote: > >> Il 10 Gen 2018 20:40, <ovirt@fateknollogee.com> ha scritto: >> >>> oVirt + Gluster (hyperconverged) RAID question: >>> I have 3 nodes of SuperMicro hardware, each node has 1x SATADOM >>> (boot drive for o/s install) and 6x 1TB SSD (to be used for >>> Gluster). >>> For the SSDs, is hardware or software RAID preferred or do I use >>> an >>> HBA? >>> The RedHat docs seem to suggest hardware RAID, others on the forum >>> say HBA or software RAID. >>> >>> What are other folks using? >> >> I'm using hardware raid. >> >>> _______________________________________________ >>> Users mailing list >>> Users@ovirt.org >>> http://lists.ovirt.org/mailman/listinfo/users [1] [1] >> >> Links: >> ------ >> [1] http://lists.ovirt.org/mailman/listinfo/users [1] > > > > Links: > ------ > [1] http://lists.ovirt.org/mailman/listinfo/users
I really don't see much need of using hardware raid if you have a SSD-only environment. You will get little benefit from hardware cache memory and to guarantee the writes you may having the filesystem doing always sync operations, similar to what ZFS does. You just need to use a HBA to controll all disks ou even software RAID will do the job. Fernando On 10/01/2018 17:34, ovirt@fateknollogee.com wrote:
oVirt + Gluster (hyperconverged) RAID question: I have 3 nodes of SuperMicro hardware, each node has 1x SATADOM (boot drive for o/s install) and 6x 1TB SSD (to be used for Gluster). For the SSDs, is hardware or software RAID preferred or do I use an HBA? The RedHat docs seem to suggest hardware RAID, others on the forum say HBA or software RAID.
What are other folks using? _______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
On Wed, Jan 10, 2018 at 10:12 PM, FERNANDO FREDIANI < fernando.frediani@upx.com> wrote:
I really don't see much need of using hardware raid if you have a SSD-only environment. You will get little benefit from hardware cache memory and to guarantee the writes you may having the filesystem doing always sync operations, similar to what ZFS does. You just need to use a HBA to controll all disks ou even software RAID will do the job.
I think there are two interesting questions here: 1. Why would you want RAID? Your data is replicated already on other Gluster nodes. RAID writes will wear out the SSDs earlier and RAID 'wastes' additional space. The reason may be because you don't want the node to go down in case of a single disk malfunction. Fair - but is that the case? 2. I think HW RAID should be higher performing, but that's about it. Y.
Fernando
On 10/01/2018 17:34, ovirt@fateknollogee.com wrote:
oVirt + Gluster (hyperconverged) RAID question: I have 3 nodes of SuperMicro hardware, each node has 1x SATADOM (boot drive for o/s install) and 6x 1TB SSD (to be used for Gluster). For the SSDs, is hardware or software RAID preferred or do I use an HBA? The RedHat docs seem to suggest hardware RAID, others on the forum say HBA or software RAID.
What are other folks using? _______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
participants (4)
-
FERNANDO FREDIANI -
ovirt＠fateknollogee.com -
Roberto Nunin -
Yaniv Kaul