Re: high availability options
Hello all, With ceph it can be done with out any issue. Take a look at ceph.com Best regards , Koby Peleg Hen -- Sent from my Android device with K-9 Mail. Please excuse my brevity.
Re: high availability options
Janos SUTO wrote: > > Dear piler-users, > > I've got a feature request for DAG-like high resiliency. > > I've designed piler to run on a single host only. However it's a valid > business demand that let the archive be fault tolerant. > > Though piler has no builtin high availability feature, something close > might be achieved. > > Setup two piler boxes (they can be even in distant datacenters, not > necessarily > on the same lan). These 2 nodes are independent, having no knowledge > of each other. > > The only difference is that the smtp copy must be done to two addresses: > archive@piler1 and archive@piler2. Thus you have an email archived at two > places. Users should access http(s)://archive which should point to > piler1, > and in case of trouble you just point it to piler2. Hi, the first issue I can note is that in this case I would have two archives, but during time the secondary copy will never be syncronized to the first one. Deleted emails for example would not be deleted on the secondary, and so on. F. > > I'm not sure if such a setup may be ok for you, but I'm interested in > your > ideas and expectations. > > Janos >
Re: high availability options
"Janos SUTO" wrote: > >The only difference is that the smtp copy must be done to two addresses: >archive@piler1 and archive@piler2. Thus you have an email archived at >two places. If I were designing this, I would set up a DNS MX record with different priorities for each Piler box. This way, making it work with any mail systems that only support a single email address for a journaling/archive mailbox (if there are any). On the back end, have Piler re-queue the message and send it via SMTP to the other cluster member(s) and then archive it. When a cluster member sees an SMTP connection from another member, it should just archive the message and not queue it up (to prevent a loop) back to the other cluster member(s). In the event of a cluster member failure, the receiving member will just keep the messages spooled via SMTP until the failed member comes back up. The only failure that could cause message loss that I can see on brief examination is if the receiving cluster crashes a disk before passing the spooled message to the other members. Just thinking out loud... -Arthur - Arthur Emerson III Email: emer...@msmc.edu Network Administrator InterNIC: AE81 Mount Saint Mary College MaBell: (845) 561-0800 Ext. 3109 330 Powell Ave.Fax:(845) 562-6762 Newburgh, NY 12550SneakerNet: Aquinas Hall Room 11
high availability options
Dear piler-users, I've got a feature request for DAG-like high resiliency. I've designed piler to run on a single host only. However it's a valid business demand that let the archive be fault tolerant. Though piler has no builtin high availability feature, something close might be achieved. Setup two piler boxes (they can be even in distant datacenters, not necessarily on the same lan). These 2 nodes are independent, having no knowledge of each other. The only difference is that the smtp copy must be done to two addresses: archive@piler1 and archive@piler2. Thus you have an email archived at two places. Users should access http(s)://archive which should point to piler1, and in case of trouble you just point it to piler2. I'm not sure if such a setup may be ok for you, but I'm interested in your ideas and expectations. Janos
