2 storage servers and a watch server
ek-media
3 Posts
April 18, 2017, 4:12 pmQuote from ek-media on April 18, 2017, 4:12 pmHi,
Is it possible to have 2 SSD storage servers (12 TB) and a third watch/monitoring server without a data storage?
Hi,
Is it possible to have 2 SSD storage servers (12 TB) and a third watch/monitoring server without a data storage?
admin
2,930 Posts
April 19, 2017, 9:02 amQuote from admin on April 19, 2017, 9:02 amYes you can un-check the "Local Storage Service" role on the node that you do not want serving storage.
More precisely, there are 3 roles a node can perform:
1) Management and Monitoring
2) Local Storage
3) iSCSI Target
The first is non-optional for the first 3 nodes.
The second and third are optional and be can unselected / selected at runtime.
Yes you can un-check the "Local Storage Service" role on the node that you do not want serving storage.
More precisely, there are 3 roles a node can perform:
1) Management and Monitoring
2) Local Storage
3) iSCSI Target
The first is non-optional for the first 3 nodes.
The second and third are optional and be can unselected / selected at runtime.
Last edited on April 19, 2017, 9:03 am · #2
lowey71
2 Posts
September 7, 2017, 12:18 amQuote from lowey71 on September 7, 2017, 12:18 amSo would it be possible to have have 2 ESXi servers, have 2 'Local Storage' nodes on each but and say the 3rd server (M&M only) as another VM?
Would assume that if a ESXi host fails, then may lose quorum? Would need to run Vmware FT to ensure the 3rd node is always available on either host (so 2 nodes are always available)?
We currently run Starwind in a 2 node arrangement and wondering if this might be a suitable replacement 🙂
So would it be possible to have have 2 ESXi servers, have 2 'Local Storage' nodes on each but and say the 3rd server (M&M only) as another VM?
Would assume that if a ESXi host fails, then may lose quorum? Would need to run Vmware FT to ensure the 3rd node is always available on either host (so 2 nodes are always available)?
We currently run Starwind in a 2 node arrangement and wondering if this might be a suitable replacement 🙂
admin
2,930 Posts
September 7, 2017, 11:10 amQuote from admin on September 7, 2017, 11:10 am
running your third monitor in FT/HA under VMWare would do, another option is running it on any third/extra ESX you have..it requires very low system resources so you can fit the vm in a low end ESX while the first 2 storage machines that house the storage would be more powerful boxes.
to setup FT/HA will require the vm to use a SAN based datastore rather than local storage. Trying to use PetaSAN as datastore for its quorum vm will probably not work.
running your third monitor in FT/HA under VMWare would do, another option is running it on any third/extra ESX you have..it requires very low system resources so you can fit the vm in a low end ESX while the first 2 storage machines that house the storage would be more powerful boxes.
to setup FT/HA will require the vm to use a SAN based datastore rather than local storage. Trying to use PetaSAN as datastore for its quorum vm will probably not work.
Last edited on September 7, 2017, 11:11 am by admin · #4
lowey71
2 Posts
September 7, 2017, 9:20 pmQuote from lowey71 on September 7, 2017, 9:20 pmThanks for the response. 2 box clusters are great for ROBO and in our case, supplying redundant services away from our primary SAN in case of failure.
If I get the chance, will test and advise.
Would imagine if the ESXi host dies that is hosting the storage for the 3rd M&M service, and that host was also providing the storage/iscsi endpoint for that VM, any disk access will fail until the paths migrate over to the other host/VM Ceph storage server.
I feel that as long as no disk access is required on that witness to complete the storage failover, then it would work 🙂
Thanks for the response. 2 box clusters are great for ROBO and in our case, supplying redundant services away from our primary SAN in case of failure.
If I get the chance, will test and advise.
Would imagine if the ESXi host dies that is hosting the storage for the 3rd M&M service, and that host was also providing the storage/iscsi endpoint for that VM, any disk access will fail until the paths migrate over to the other host/VM Ceph storage server.
I feel that as long as no disk access is required on that witness to complete the storage failover, then it would work 🙂
2 storage servers and a watch server
ek-media
3 Posts
Quote from ek-media on April 18, 2017, 4:12 pmHi,
Is it possible to have 2 SSD storage servers (12 TB) and a third watch/monitoring server without a data storage?
Hi,
Is it possible to have 2 SSD storage servers (12 TB) and a third watch/monitoring server without a data storage?
admin
2,930 Posts
Quote from admin on April 19, 2017, 9:02 amYes you can un-check the "Local Storage Service" role on the node that you do not want serving storage.
More precisely, there are 3 roles a node can perform:
1) Management and Monitoring
2) Local Storage
3) iSCSI Target
The first is non-optional for the first 3 nodes.
The second and third are optional and be can unselected / selected at runtime.
Yes you can un-check the "Local Storage Service" role on the node that you do not want serving storage.
More precisely, there are 3 roles a node can perform:
1) Management and Monitoring
2) Local Storage
3) iSCSI Target
The first is non-optional for the first 3 nodes.
The second and third are optional and be can unselected / selected at runtime.
lowey71
2 Posts
Quote from lowey71 on September 7, 2017, 12:18 amSo would it be possible to have have 2 ESXi servers, have 2 'Local Storage' nodes on each but and say the 3rd server (M&M only) as another VM?
Would assume that if a ESXi host fails, then may lose quorum? Would need to run Vmware FT to ensure the 3rd node is always available on either host (so 2 nodes are always available)?
We currently run Starwind in a 2 node arrangement and wondering if this might be a suitable replacement 🙂
So would it be possible to have have 2 ESXi servers, have 2 'Local Storage' nodes on each but and say the 3rd server (M&M only) as another VM?
Would assume that if a ESXi host fails, then may lose quorum? Would need to run Vmware FT to ensure the 3rd node is always available on either host (so 2 nodes are always available)?
We currently run Starwind in a 2 node arrangement and wondering if this might be a suitable replacement 🙂
admin
2,930 Posts
Quote from admin on September 7, 2017, 11:10 am
running your third monitor in FT/HA under VMWare would do, another option is running it on any third/extra ESX you have..it requires very low system resources so you can fit the vm in a low end ESX while the first 2 storage machines that house the storage would be more powerful boxes.
to setup FT/HA will require the vm to use a SAN based datastore rather than local storage. Trying to use PetaSAN as datastore for its quorum vm will probably not work.
running your third monitor in FT/HA under VMWare would do, another option is running it on any third/extra ESX you have..it requires very low system resources so you can fit the vm in a low end ESX while the first 2 storage machines that house the storage would be more powerful boxes.
to setup FT/HA will require the vm to use a SAN based datastore rather than local storage. Trying to use PetaSAN as datastore for its quorum vm will probably not work.
lowey71
2 Posts
Quote from lowey71 on September 7, 2017, 9:20 pmThanks for the response. 2 box clusters are great for ROBO and in our case, supplying redundant services away from our primary SAN in case of failure.
If I get the chance, will test and advise.
Would imagine if the ESXi host dies that is hosting the storage for the 3rd M&M service, and that host was also providing the storage/iscsi endpoint for that VM, any disk access will fail until the paths migrate over to the other host/VM Ceph storage server.
I feel that as long as no disk access is required on that witness to complete the storage failover, then it would work 🙂
Thanks for the response. 2 box clusters are great for ROBO and in our case, supplying redundant services away from our primary SAN in case of failure.
If I get the chance, will test and advise.
Would imagine if the ESXi host dies that is hosting the storage for the 3rd M&M service, and that host was also providing the storage/iscsi endpoint for that VM, any disk access will fail until the paths migrate over to the other host/VM Ceph storage server.
I feel that as long as no disk access is required on that witness to complete the storage failover, then it would work 🙂