Full sync after restart

[logitech]

Hello everyone,

I'm facing an issue with StarWind software that I hope someone can help with. After restarting, I noticed a full sync job automatically started for the nodes. This process is taking an unusually long time—several hours.

This delay is causing significant downtime and disruption to our operations. I've attached the log below for more details.

Has anyone experienced a similar issue, or does anyone have suggestions on preventing this from happening in the future? Any help would be greatly appreciated!

[yaroslav (staff)]

Hi,

I need logs from both nodes.

After restarting, I noticed a full sync job automatically started for the nodes.

The short answer is to let it synchronize. Just plan the time slot where you can finish the sync (e.g., weekend). Until then, keep the not synchronized node down.
Full sync after the restart can happen because of mishandling the restart procedure or write-back cache. See more https://knowledgebase.starwindsoftware. ... may-start/.
See how to remove the cache https://knowledgebase.starwindsoftware. ... -l1-cache/.

Could you please let me know what the storage configuration is (e.g., RAID level)?
Reduce synchronization priority. You will need to stop the StarWindService on both sides and modify <sync_traffic_share>50</sync_traffic_share> to 5 for each *_HA.swdsk
Start the service afterward.

[logitech]

Hi Yaroslav,

Well noted for the instructions. much appreciated

I have attached the service logs for both nodes and we are using RAID5 on SSDs mix-load.

Link for the attachment:
https://we.tl/t-yEvKMkKn2F

Cheers!!!

[yaroslav (staff)]

You are welcome
Could you please share the standard StarWind support bundles?
For Windows-based app, check https://knowledgebase.starwindsoftware. ... collector/
For CVM, Press the gear-shaped button -> Expand the VM name in the first tab -> collect support bundle.

[logitech]

Morning Yaroslav!

The logs below in the link

https://we.tl/t-HUzBbu5D58

I am awaiting your prompt reply.

[yaroslav (staff)]

The cause for full synchronization is the simultaneous shutdown of both nodes 8/19/2024 8:59:12 PM. I can also see that the synchronization is over.
Next time you can change synchronization priority with haSyncPriority.ps1 to change the synchronization priority of the devices and let synchronization finish with minimum impact on production. Set it to ~5%.
And make sure to follow the recommended restart procedure https://knowledgebase.starwindsoftware. ... installed/

[logitech]

Hi Yarsolav!!

Thanks for the update.

I'm currently facing an issue with my SQL Server VM where the performance is noticeably slow, especially during certain queries. I've observed that the disk utilization spikes to 100% during these operations, which is causing significant delays.

sql.jpg (186.48 KiB) Viewed 10527 times

I suspect that there might be some configuration settings that need to be fine-tuned to optimize the performance. Has anyone experienced a similar issue or can you suggest specific adjustments that could help mitigate this problem?

Any advice or recommendations would be greatly appreciated!

[yaroslav (staff)]

Thanks for your update. Just general ones:
Does the VM have any snapshots?
Is its disk fixed?
Did it start happening after the outage?
Could you try enabling a write-back cache for the RAID controller?
Try shutting down the VM and playing around with CPU allocation (NUMA, Hardware topology, etc).
Make sure the NIC drivers are up-to-date.
Try assigning more local iSCSI sessions for StarWind HA devices.
Make sure MPIO is set to LQD on both nodes.

[logitech]

Does the VM have any snapshots?
NOP
Is its disk fixed?
they are thin virtual disks
Did it start happening after the outage?
I don't think so
Could you try enabling a write-back cache for the RAID controller?
I will update the configuration
Try shutting down the VM and playing around with CPU allocation (NUMA, Hardware topology, etc).
I need to do some research about it since I don't know what to do exactly here, I have assigned 16 cores for this VM.
Make sure the NIC drivers are up-to-date.
They are up-to-date.
Try assigning more local iSCSI sessions for StarWind HA devices.
I'm using the MS Windows iSCSI initiator (2 sessions), do you recommend using Starwind iSCSI Accelerator? Is it straightforward forward to install it? does it need to restart after the installation?
Make sure MPIO is set to LQD on both nodes.
The least Queue depth is already configured.

[yaroslav (staff)]

Try thick disks, enable write back controller cache, and play around the CPU allocation.
You can add more loopback sessions as described here https://www.starwindsoftware.com/resour ... c3bac25222. This does not need downtime. Do it for CSVs.

[logitech]

Quick update:

We changed the RAID configuration from RAID5 to RAID10, and SQL performs very well.

Changes made without any impact:

- Enabled write-back cache on the raid controller and played with the cache ratio 60% read 40% write / 50% read 50% write. no luck!
- Added more sessions also no luck!

Last question about Starwind iSCSI Accelerator, does it require a system restart to take effect?

[yaroslav (staff)]

Great news.
Adding more iSCSI sessions does not need an OS restart.