Hello

 

This evening I have added a couple of Samsung 500gb ssd’s to our Unraid server, which is an Hp ML310e g8 V2. The drives are currently connected to the two on board sata connectors. The server had no cache before this. 
 

When using only one ssd as a cache disk I can write to the server at a constant 113mbs.
 

When adding the second ssd to create a cache pool, the results are pretty poor. 
 

It will copy at 113mbs for about 20seconds, it will then drop right down to less than 1mbs. At this point Unraid shows that we’re writing to one of the cache drives, then alternates between the two. 
 

Using Winscp to write to it, pops up saying the server isn’t responding, but continues at super slow speed. ~200kbs

 

Tried both ssd’s separately as cache and can max out the 1gb Ethernet all day long. 
 

Have ordered a sata to mini sas cable to try the mother board sas connector and port2 on the HBA card to see if it makes any difference with both disks in a cache pool. 
 

Has anyone seen similar issues before with more than one cache drive?

 

Many thanks. 

When you add a new SSD to a pool, btrfs balance will automatically run which redistribute data to the 2 disks depending on what raid scheme you pick (default is RAID-1). Did you wait for balance to finish before testing?

 

You are mixing byte with bit. Small b is bit, capital B is byte.

 

If you are only accessing the server through the network, you don't really need cache since turbo write alone in most cases would saturate gigabit network.

 

Thanks for the response. I typed that post late last night on my phone which is why the b and B's are all little.

Its running in default raid 1 and yes I waited for the balance to complete each time before testing.

 

I might only be accessing it via the network at the moment, but there is something wrong with the hardware or software for the speed to drop off so badly after 20/30secs. It goes from 113MB/s to less than 1MB/s and then down to below 0.2MB/s it never picks back up.

 

This doesn't happen gradually, it drops straight down to less than 1MB/s

 

I have multiple dockers i would like to move the cache and a VM. It would seem they might not run very well with performance like that.

PS I have 10gbe waiting to be installed too, which was waiting on the cache drives.

1 hour ago, dansonamission said:

I might only be accessing it via the network at the moment, but there is something wrong with the hardware or software for the speed to drop off so badly after 20/30secs. It goes from 113MB/s to less than 1MB/s and then down to below 0.2MB/s it never picks back up.

Perhaps, this was the issue

 

Thanks, just read the entire thread. Seems it’s been going on for years with Samsung drives and no one who could do anything about it seems to care. 
 

I’ll check the performance of the rest of the system when the transfer speed drops. I know winscp complains of connection issues when transferring large files to the cache pool.