Optimal guest os cache settings question

[Romsla]

Hi.
Still using XFS NFS NAS,but re-evaluating starwind LSFS. A while ago (around half a year) I decided not to use starwind software as it had some issues.

My setup and usage goal:
A single battery-backed write-back hardware raid 6 local hdd sas storage on single vmware esxi 5.5 host. (esxi does no caching at all and guarantees the read\write consistency on hypervisor level)
The plan is to utilize LSFS to leverage better write perfomance and dedupe for backups residing on the guest Windows OS.
Starwind LSFS will be installed on Windows guest OS. The same Windows guest OS will contain the backup system. However it could be installed on separate guest OS if requred.
The optional goal is to leave write-back enabled on hardware raid utilized for LSFS, as it is battery-backed and should not be an issue here.
The main goal, and It is the critical one - in the event of a sudden power failure and\or any single-sector SAS timeout\error LSFS must not be corrupted and under any circumstances must not be in a state requiring any paid support recovery procedures, however any pending NTFS writes to LSFS volume are allowed to be lost.

So the questions are:
1.Can I use the same guest OS for backup and lsfs hosting the ntfs volume used for them?
2.What guest OS and starwind lsfs cache settings should I use? It is at all possible to gurantee LSFS consistency in such scenario? (please also consider that windows won't allow to turn off write cache on virtualized disks in some scenarios)

Thank you.

[Romsla]

Another way to put my question:
Can you guarantee the instant recoverability of LSFS volume in case of power failure and\or single-sector disk error with acceptable data loss of data in question of being written?
Is it good idea to use LSFS for backups at ALL ???
How does LSFS handle disk errors and power failures?

The reason for my quesition is very simple really, considering the following scenario:

The power fails, and a single production VM gets corrupted (for example corruption due to guest os writes not flushed). If backup system which is LSFS backed is corrupted (lsfs volume in need of paid support recovery that may take days), than it is not acceptable.
As an example - XFS volume will be accessible in any case of single-sector error or power failure. The data loss may affect the data, but never the volume and will not spread to adjacent data.

So basically...
Some Data loss - OK. Volume loss due to data loss - unacceptable. Should I use LSFS for backup storage yes\no question? If Yes- what Cache settings on guest os and lsfs?

PS UPS is online, but power failure may be due to power grid overload or short circuit, as it it a datacenter with shared rack space.

[darklight]

Hi Romsla,

Standalone LSFS device with deduplication (will eat some CPU and RAM) and without any caching should be fine. I am using exactly the same configuration to store installation distributives and share them into the corporate network over SMB.

[Tarass (Staff)]

As already stated by my colleagues in other threads LSFS in latest build is ready for production usage. Make sure you have downloaded the latest one from your personal page on our website.