Probable data loss bug

[aulevalar]

Good day,

So far I'm extremely impressed with the beta caching software. I'm looking forward to the ability to have persistency of caching using a SSD so that my large RAID array is speed optimized even during the first run of an application.

I have run across a configuration that has caused data loss. I would like to take a guess at the root cause of the problem, as I'm the prototypical geek resulting from picking up a degree in electrical engineering (not that all EE's end up as geeky as myself *smile*). Here is my current configuration:

Fancy Cache for volumes set as active on drives C: and E:.
Boot Drive C: 120GB SSD, set to 512MB defered write RAM cache
Data Drive E: 1.5TB Raid level 5, set to 2048MB defered write RAM cache, Level 2 cache of 20GB on drive C: (SSD)

I think the problem is that with the level 2 cache as the SSD drive, which itself has a RAM cache, then the process goes from RAM over E: to RAM over C: finally to SSD physical them to the final disk platters on E:.

I disabled the RAM cache on drive C: to resolve this issue, but I thought that your team should know about this problem of having essentially 3 tiers of caching. I would think that some form of solution to this would include the ability to defer write cache to the SSD to reduce the amount of small data writes. My solution of turning off the cache on the SSD is not the preferred solution.

Looking forward to any help I may be able to offer,
Aule

[Support]

Hi aulevalar,

Thank you for your feedback.

Yes, you're right. Such settings (tied or linked caching) may cause unexpected results. I think it is not easy to work out a perfect solution. Actually such scenarios had been considered at the intial design phase. But finally we rejected it because of the technical issues and besides it will introduce lot of risks.