tucansam Posted September 22, 2012 Share Posted September 22, 2012 I have a 320GB 2.5" disk lying around that I'd like to use as my cache drive. I only have room in my server for a 2.5" drive (two, actually). If I configure my MB's onboard SATA to run a pair of 2.5" disks in RAID0, this would help with my space issues. However, is there something that happens when the cache drive fills before the mover script runs, or does the system simply return disk full errors? I've thought about making the mover script run once per hour. Or just creating a script that would run it manually, as the only time this would be a real problem would be during my initial data copy (have some 5TB of data to copy over). Advice? Link to comment
BetaQuasi Posted September 23, 2012 Share Posted September 23, 2012 Do the initial transfer prior to adding the cache drive. Link to comment
Influencer Posted September 23, 2012 Share Posted September 23, 2012 Yes do the transfer before adding the cache. Using the cache will cause your data to be unprotected until transferred to the array. If you do not mind the temporary lapse in protection you can disable your parity until the 5 tb is transferred, this will have the same effect as using the cache drive without having to move the data twice. Link to comment
mrow Posted September 25, 2012 Share Posted September 25, 2012 I agree with the rest. Invoking the mover once an hour is still going to fill the cache up faster than it can move that data to the array considering over a gigabit network you will be writing to the cache drive at around 3 times faster than the data can be moved to the array. Just disable the cache drive for now. Link to comment
lionelhutz Posted September 25, 2012 Share Posted September 25, 2012 Running the mover while moving files will just slow down the access to the drive. Platter drives don't like to seek. There is a free space setting for the cache. Set it and then files will copy to cache until you reach the min free space you set. Once you do that, the files go directly to the array. Link to comment
methanoid Posted January 7, 2016 Share Posted January 7, 2016 This was the thread I came to when looking for info about how the cache works (or maybe SHOULD work). What I would like to see is that the cache is allowed to fill to a certain % before being bypassed (effectively disabling cache) AND the mover script auto-run...before the cache is re-enabled.. Link to comment
HellDiverUK Posted January 7, 2016 Share Posted January 7, 2016 I just turn off caching for whatever share I'm populating, then turn it back on afterwards. Don't have to stop the array to do it either. Link to comment
methanoid Posted January 7, 2016 Share Posted January 7, 2016 Sorry, I was talking more about how I would LIKE the cache to behave and the mover script.... but yes, good advice for OP (but somewhat passed the sell by date!) Link to comment
trurl Posted January 7, 2016 Share Posted January 7, 2016 This was the thread I came to when looking for info about how the cache works (or maybe SHOULD work). What I would like to see is that the cache is allowed to fill to a certain % before being bypassed (effectively disabling cache) AND the mover script auto-run...before the cache is re-enabled.. Absolute size rather than percentage for Min Free Space is the way it currently works and that is the way it should work. You set Min Free Space to be larger than the largest file you expect to copy . When you copy a file that would make it go below that setting it goes to the array instead, so you never run out of cache space when copying to user shares. Note that cache is not actually disabled by Min Free Space. You could have min free set to 50GB and when it gets to 55GB, if you try to copy a 10GB file it will go to the array instead, but you can still copy a 2GB file to cache. Running mover when copying is probably counter-productive so that is probably why it has never been automatically triggered. Link to comment
lionelhutz Posted January 8, 2016 Share Posted January 8, 2016 This was the thread I came to when looking for info about how the cache works (or maybe SHOULD work). What I would like to see is that the cache is allowed to fill to a certain % before being bypassed (effectively disabling cache) AND the mover script auto-run...before the cache is re-enabled.. And as I answered before, writing to the array while the mover is working will really slow down the write operation. Trying a large read while the mover is writing to the same disk would also slow things down. Hence the current method of filling the cache and then just writing to the array. The mover running when scheduled allows you to do it when you're not using the array so it doesn't slow anything down. What happens if you don't fill the cache enough to trigger the mover for quite a while? Will the data just sit there for days or weeks until enough data is written? Asking for a feature - that gets done in the feature request section. https://lime-technology.com/forum/index.php?board=63.0 Link to comment
Recommended Posts
Archived
This topic is now archived and is closed to further replies.