I’ve read a number of reports over the past couple of days that indicate the demise of Project Wonderland and Open Solaris. As it is open source I would not be surprised to see the project continue, even if forked and retitled in some way. Oracle will continue selling Solaris, but “nothing is for free” is about to take on a whole new meaning. For now it seems that any plans around packaging services around the FOSS stack surrounding former Sun products should be put on hold until further official announcements are made.
“So now, less than five years later, you can go up on a steep hill in Las Vegas and look west, and with the right kind of eyes you can almost see the high-water mark – that place where the wave finally broke and rolled back.”
–Hunter S. Thompson
Goodbye Sun. You will be sorely missed.
Moving on. For what it is worth I’m in the process of migrating the non-Mac SAN portion of my home lab to Debian with a little OpenDedup and OpenFiler until I can decide which to centralize my VMs onto. For now I’m mostly looking to see how much deduplication occurs with the VMs. I ran into an issue where the VMs were unstable with my OpenDedup volume. I was able to fix this by disabling deduplication. To do so, you edit the VOLUMENAME-volume-cfg.xml file (transposing your volume name for VOLUMENAME), editing the dedup-files=”true” to read dedup-files=”false”.
Turned out that my issue was in part a latency issue with the location of the deduplication database, which I then mv’d to an SSD drive and edited the dedup-db-store and meta-db-store to point at the new location. Then I re-enabled dedup-files and my VMs were working again. I also ended up increasing my max-open-files to 2048, also in that same file. The other issue was with the fact that I was doing inline deduplication. To fix that, I switched to batch deduplication. According to the documentation this is preferred for most environments that aren’t backup targets.
Overall I’ve been pretty happy with OpenDedup and really only just begun testing OpenFiler. In my initial tests I’m very satisfied with the block level replication, the ease with which you create iSCSI targets and snapshots. I’m actually finding the transition to OpenFiler easier than the learning curve was with OpenStorage (how hard can anything with a GUI really be?), although I am intrigued by the savings with OpenDedup and given its learning curve, continue to tinker when I can.
Final note, if you want to run a snapshot of your OpenDedup volume, use the following (assuming your data destination is /my/destination and the source location that you will be creating a snapshot of is /my/source) then use or script the following to run on timed intervals:
setfattr -n user.cmd.snapshot -v 5555:/my/destination /my/source
When I did this there was a slight amount of latency created, although that might be less noticeable when used on better hardware…