> On Mar 8, 2019, at 8:07 PM, Benjamin Kaduk <[email protected]> wrote: > > On Tue, Mar 05, 2019 at 10:25:23PM +0100, Måns Nilsson wrote: >> Subject: [OpenAFS] About `dafileserver` vs `fileserver` differences (for >> small cells) Date: Tue, Mar 05, 2019 at 10:05:25PM +0200 Quoting Ciprian >> Dorin Craciun ([email protected]): >>> Hello all! >>> >>> I understand from the documentation that the main difference between >>> `dafileserver` and `fileserver` is the "on-demand-attach" of volumes. >> >> <snip> >> >>> I ask this also from the perspective of a small cell operator (for >>> personal purposes), where attach-on-demand is not an issue, and in >>> fact I think I would prefer all my volumes to be attached as early as >>> possible. >> >> As I recall, the demand attach file server is a scalability option. Which in >> practice means that if you need it, you know it. Me, I have a startup time >> under one minute for my small cell of some 400 volumes over 1,5TiB data. >> This using a small and old Supermicro machine with ZFS and FreeBSD. > > My recollection was that the dafs was to be generally preferred in all > cases. But I haven't interacted closely with those bits, so I'm just > relaying hearsay; it would be good if someone closer to that work could > chime in.
DAFS main benefit is the reduced impact of restarting a fileserver, especially fileserver with thousands or even millions of volumes. DAFS fileservers are able to restart more quickly, are able to avoid restarts formerly required for volume salvages, and are able to reduce the negative effects of restarts on clients. Here are some details about how these benefits are acheived: 1) As the name Demand Attach implies, volumes are only attached on demand - that is, when referenced by clients - and detached automatically after a short period of non-reference (2hr default). This eliminates the startup delay incurred on the old fileserver while attaching all volumes before any could be served. It may also greatly reduce the resources required to manage lots of volumes, because only volumes that are in active or recent use are under active management by the fileserver. 2) Volumes may be salvaged without a fileserver outage; in fact, the fileserver will automatically perform volume salvages when damage is detected, without any administrator intervention required. 3) Full partition salvages still require an outage. However, DAFS is able to run these in parallel in order to reduce the length of the outage. 4) The fileserver saves the state of all active hosts and all callbacks at shutdown. If restart occurs within 30 minutes, this state file is restored so that callbacks don't need to be revoked; thus clients don't need to prematurely refetch status or data. Regards, -- Mark Vitale [email protected]
