Re: HAMMER2 PFSes
Right, I wasn't trying to use the clustering feature. "cluster" appears in the docs and a lot of the commands take or show a cluster ID, so I wondered what it meant it the system as it stands now. I might have been trying to read it too literally. Also, the formatting of my earlier questions got screwed up and were hard to read. I'll try again. If you create a master PFS on a filesystem without specifying a cluster ID, I think what you end up with is an entirely independent "root" that can use whatever space is available on the filesystem. It's independent of other PFSes, meaning that the content is different. Is that a correct understanding? If you create a master on that filesystem and specify the cluster ID that already exists, I'm not sure what you end up with. It does allow that, however. What is that configuration (multiple masters for a cluster ID) for? Maybe that would useful for a cluster distributed across machines, but not currently? If you create a slave (by specifying an existing cluster ID), it seems to be what you'd expect: a copy of the master. But you can't mount it when the master is also mounted. Did I do something that doesn't make sense? As for the cluster ID, I didn't consider the possibility that you should just pay attention to it or use it at all in the commands. Maybe I should just not use it now? Chuck On 2020-04-23 19:49,atthew Dillon wrote: > For now don't try to cluster anything. That work is still in-progress. You > can create multiple independent masters on the same device, and you can > snapshot them, as well as be able to write to the snapshots. That all works. > The snapshots basically work the same as masters. > > -Matt
Re: HAMMER2 PFSes
For now don't try to cluster anything. That work is still in-progress. You can create multiple independent masters on the same device, and you can snapshot them, as well as be able to write to the snapshots. That all works. The snapshots basically work the same as masters. -Matt
HAMMER2 PFSes
I'm trying to understand some of the concepts behind HAMMER2, so I played with a freshly installed 5.8 system and have a few questions. 1. As I understand it, PFSes allow a HAMMER2 filesystem to be split into isolated parts which share the space available of the underlying filesystem. Looking closer at the manual, I'm wondering if the actual "unit of separation" is really the entity called a cluster. Is that an accurate assessment? The hammer2 manpage mentions cluster frequently, but I couldn't find a definition for it. Last I heard, the multi-machine clustering feature isn't implemented yet, so clusters are inherently local to a machine currently. 2. I created a MASTER PFS in the main filesystem of a fresh 5.8 system, put a file in it and then created a SLAVE PFS, specifying the cluster ID of the MASTER I'd just made. My understanding was the SLAVE was a read-only mirror of the MASTER. I tried to mount the MASTER, then the SLAVE (at two different mount points of course), but the SLAVE mount attempt failed with "device busy". When mounting just one at a time, they contain the same file (as I'd expect), and modifications to it show up when the other partition is mounted (after dismounting the one I was working with). What does "busy" mean in this case? What's the proper usage for a SLAVE PFS? I figured it was to have read-only duplicates, but the way I tried it didn't work. 3. I created a snapshot of the MASTER PFS I'd made earlier and it shows up as a different kind of PFS (a SNAPSHOT). What is the difference between creating a snapshot and creating a MASTER or SLAVE in an already existing cluster? 4. I created an additional MASTER for a cluster that already has a MASTER. It let me do that, but what are the implications of doing so? Thanks, Chuck