- This topic has 6 replies, 4 voices, and was last updated 11 months, 2 weeks ago by
.
-
Topic
-
-
Replies
-
-
-
Good morning Peter,
The biggest challenge looks like over coming the fact that you data disk isn’t replicated (that’s how I am reading the attached) which means that an organization has to either accept
- at an automatic failover your data disks will not be in sync (i.e. smat, logs etc) which could be an issue if you want to validate no data loss
- failovers are not automatic, but manual events where your data disk is detached from the passive and moved to the soon to be active server
- put a strategy in place to keep you data disks in sync (i.e. when a message is received, before it is ack’d back to the sender push it over to to your passive node and when you get that ack, you ack back to the sender)
Doesn’t seem like any of these solutions are ideal.
-
-
We have been ‘ordered’ to build a completely new Cloverleaf (TST and PRD where each is an active/passive HA setup) in Azure. I’m sure there are a few ways to do this; the most normal one is building 4 VM RHEL servers with (two) shared disks. Basically the way we build Cloverleaf on-premis.
If shared disks are not possible in Azure (why not?) then this would be a no-go for me. If you need to send messages to the other (passive) server as well, you can still miss messages in case of a failover.
Our hospital wants to use its ‘own’ Azure environment, so the AWS Cloverleaf in cloud is not an option at the moment. But very interested to hear what other Cloverleaf users are doing in cloud.
Zuyderland Medisch Centrum; Heerlen/Sittard; The Netherlands
-
I tested the performance of Azure shared disks about a year or so ago. All of them were too slow for Cloverleaf (or had other showstopper issues) except for the Azure Premium shared disk. The premium shared disk was fast enough for Cloverleaf. Things may have changed since then, so take this information with a grain of salt.
If you would like to do some testing on your own, I’ll share the Cloverleaf site I used for performance testing. Shoot me an email if you need it.
-- Max Drown (Infor)
-
I found this in my notes for Azure shared disk testing from Oct. 14, 2021.
CISF: doesn’t support FIFO queues, cannot use this filesystem.
NFS: TPS is slow, cannot handle the volume.
Again note that the technology may have changed since I did this testing.
-- Max Drown (Infor)
-
Thanks for the information, Max. Will certainly keep an eye on those shared disks.
Then there is also the Docker option. Not sure if that is usable for us in a production environment. Do I have full control (for crontab, samba-mounts) of the OS in a Docker? And how does HA work then?
Lots of questions; let’s see where this takes us. Will keep you informed.Zuyderland Medisch Centrum; Heerlen/Sittard; The Netherlands
-
- You must be logged in to reply to this topic.
