Hi Folks, Anybody have any experience and/or opinions re. Remus vs. Pacemaker/DRBD as a high-availability solution for a very small cluster (2 machines)? Thanks, Miles Fidelman -- In theory, there is no difference between theory and practice. In<fnord> practice, there is. .... Yogi Berra _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
It''s worth noting that these are slightly different solutions. Remus keeps a running (but paused), completely synchronized version of your domU(s), and, when it senses that the original domU is not available, unpauses the standby domU from it''s last synchronized state. This is really any extension of live migration, as remus is just continually live migrating your memory and disk data, but never shuts down the original or unpauses the standby version unless something happens to the original. Pacemaker/DRBD, on the other hand, just synchronizes the data, and if it detects that one of the domUs has died, starts it up somewhere else. So, with Remus, the theory/goal is 0 downtime of your domU, whereas Pacemaker simply minimizes downtime to a certain point - the time it takes to detect failure and boot the new domU. -Nick>>> On 2010/09/06 at 15:40, Miles Fidelman <mfidelman@meetinghouse.net> wrote: > Hi Folks, > > Anybody have any experience and/or opinions re. Remus vs. Pacemaker/DRBD > as a high-availability solution for a very small cluster (2 machines)? > > Thanks, > > Miles Fidelman-------- This e-mail may contain confidential and privileged material for the sole use of the intended recipient. If this email is not intended for you, or you are not responsible for the delivery of this message to the intended recipient, please note that this message may contain SEAKR Engineering (SEAKR) Privileged/Proprietary Information. In such a case, you are strictly prohibited from downloading, photocopying, distributing or otherwise using this message, its contents or attachments in any way. If you have received this message in error, please notify us immediately by replying to this e-mail and delete the message from your mailbox. Information contained in this message that does not relate to the business of SEAKR is neither endorsed by nor attributable to SEAKR. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Nick Couchman wrote:> It''s worth noting that these are slightly different solutions. Remus keeps a running (but paused), completely synchronized version of your domU(s), and, when it senses that the original domU is not available, unpauses the standby domU from it''s last synchronized state. This is really any extension of live migration, as remus is just continually live migrating your memory and disk data, but never shuts down the original or unpauses the standby version unless something happens to the original. Pacemaker/DRBD, on the other hand, just synchronizes the data, and if it detects that one of the domUs has died, starts it up somewhere else. So, with Remus, the theory/goal is 0 downtime of your domU, whereas Pacemaker simply minimizes downtime to a certain point - the time it takes to detect failure and boot the new domU. > >Understood. In my case, I''m running Pacemaker/DRBD and can live with short outages. But... it''s pretty tricky to keep all the pieces configured and running properly, and I''m not looking forward to the gotchas when it''s time to migrate from Debian Lenny to Debian Squeeze - what with all different versions, dependencies, and such as one updates first one machine, then another, while trying to stay live. Remus looks a lot simpler, at the cost of the resources associated with running backup VMs. But then again, Pacemaker/DRBD requires that you have the resources to migrate to - so you need a hot spare machine anyway. Whether the backup VM is running or not, you need the same resources - and I assume the network load is similar whether its generated by Remus or by DRBD. Which comes back to the question of whether Remus is mature enough to rely on or not. Thanks, Miles -- In theory, there is no difference between theory and practice. In<fnord> practice, there is. .... Yogi Berra _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Tue, 2010-09-07 at 09:28 -0600, Nick Couchman wrote:> Pacemaker/DRBD, on the other hand, just synchronizes the data, > and if it detects that one of the domUs has died, starts it up somewhere else. > So, with Remus, the theory/goal is 0 downtime of your domU, whereas Pacemaker > simply minimizes downtime to a certain point - the time it takes to detect failure > and boot the new domU.Actually, pacemaker can do live migrations with some limitations. If the server on which the domU is running suddenly crashes, then domU does have to be booted from scratch on the other server. But if one server is taken down gracefully, Pacemaker can live migrate the domU (not instantaneous but only takes a few seconds). This does require that the domU disk images reside on shared storage though (NFS or a clustered file system like GFS or OCFS). I have no experience with Remus, but from Nick''s description of it, it sounds like Remus might be a whole lot easier to set up and may be a good way to go if all you care about is failover for your domU''s, or if you really need instantaneous failover in the event of a server crash. Pacemaker provides more flexibility in that you can have other things like DHCP/DNS servers, NFS mounts, etc. also fail over. But Pacemaker also has a very steep learning curve. Starting from scratch and never having used either Xen or Pacemaker before, it took me almost six months to build a fully functioning high availability setup for running Xen domU''s. I don''t know how long it would take to learn Remus but it sounds like it would be considerably easier. --Greg _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> > Actually, pacemaker can do live migrations with some limitations. If the > server on which the domU is running suddenly crashes, then domU does > have to be booted from scratch on the other server. But if one server is > taken down gracefully, Pacemaker can live migrate the domU (not > instantaneous but only takes a few seconds). This does require that the > domU disk images reside on shared storage though (NFS or a clustered > file system like GFS or OCFS).Right...my point was not to say that Pacemaker cannot do live migrations, but that Pacemaker does not keep a paused, up-to-date copy of the domU running somewhere else. So, if a domU crashes, or becomes unavailable for some reason, Pacemaker has to start from scratch, so there is downtime. Many folks can live with that, especially since, with decent server hardware, boot time of a domU is on the order of 10-30 seconds, depending on the services running on the domU. Others, however, want zero downtime - Remus is designed to cover these situations. -Nick -------- This e-mail may contain confidential and privileged material for the sole use of the intended recipient. If this email is not intended for you, or you are not responsible for the delivery of this message to the intended recipient, please note that this message may contain SEAKR Engineering (SEAKR) Privileged/Proprietary Information. In such a case, you are strictly prohibited from downloading, photocopying, distributing or otherwise using this message, its contents or attachments in any way. If you have received this message in error, please notify us immediately by replying to this e-mail and delete the message from your mailbox. Information contained in this message that does not relate to the business of SEAKR is neither endorsed by nor attributable to SEAKR. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Greg Woods wrote:> On Tue, 2010-09-07 at 09:28 -0600, Nick Couchman wrote: > >> Pacemaker/DRBD, on the other hand, just synchronizes the data, >> and if it detects that one of the domUs has died, starts it up somewhere else. >> So, with Remus, the theory/goal is 0 downtime of your domU, whereas Pacemaker >> simply minimizes downtime to a certain point - the time it takes to detect failure >> and boot the new domU. >> > Actually, pacemaker can do live migrations with some limitations. If the > > I have no experience with Remus, but from Nick''s description of it, it > sounds like Remus might be a whole lot easier to set up and may be a > good way to go if all you care about is failover for your domU''s, or if > you really need instantaneous failover in the event of a server crash. >All of this is well and good, but not to the point. I''m using pacemaker and DRBD. It works. it''s just a pain. Has anybody on this list actually used Remus in production? Is it ready for prime time? Or is it still beta (or alpha)? Miles -- In theory, there is no difference between theory and practice. In<fnord> practice, there is. .... Yogi Berra _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
I too would like to hear from anyone that has xen and remus in a production environment. -- Donny B. On Tuesday, September 07, 2010 11:46 AM CDT, Miles Fidelman <mfidelman@meetinghouse.net> wrote:> Greg Woods wrote: > > On Tue, 2010-09-07 at 09:28 -0600, Nick Couchman wrote: > > > >> Pacemaker/DRBD, on the other hand, just synchronizes the data, > >> and if it detects that one of the domUs has died, starts it up somewhere else. > >> So, with Remus, the theory/goal is 0 downtime of your domU, whereas Pacemaker > >> simply minimizes downtime to a certain point - the time it takes to detect failure > >> and boot the new domU. > >> > > Actually, pacemaker can do live migrations with some limitations. If the > > > > I have no experience with Remus, but from Nick''s description of it, it > > sounds like Remus might be a whole lot easier to set up and may be a > > good way to go if all you care about is failover for your domU''s, or if > > you really need instantaneous failover in the event of a server crash. > > > All of this is well and good, but not to the point. > > I''m using pacemaker and DRBD. It works. it''s just a pain. > > Has anybody on this list actually used Remus in production? Is it ready > for prime time? Or is it still beta (or alpha)? > > Miles > > -- > In theory, there is no difference between theory and practice. > In<fnord> practice, there is. .... Yogi Berra > > > > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xensource.com > http://lists.xensource.com/xen-users_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
I am also interested in a "how-to" for Remus on Xen 3.4.2 / CentOS 5.5 for Dom0. (RPMs from http://www.gitco.de/repo/GITCO-XEN3.4.2_x86_64.repo ) Any info or directions will be highly appreciated. fpt -----Original Message----- From: xen-users-bounces@lists.xensource.com [mailto:xen-users-bounces@lists.xensource.com] On Behalf Of Donny Brooks Sent: Tuesday, September 07, 2010 2:02 PM To: xen-users@lists.xensource.com Subject: Re: [Xen-users] remus vs. pacemaker/drbd? I too would like to hear from anyone that has xen and remus in a production environment. -- Donny B. On Tuesday, September 07, 2010 11:46 AM CDT, Miles Fidelman <mfidelman@meetinghouse.net> wrote:> Greg Woods wrote: > > On Tue, 2010-09-07 at 09:28 -0600, Nick Couchman wrote: > > > >> Pacemaker/DRBD, on the other hand, just synchronizes the data, > >> and if it detects that one of the domUs has died, starts it upsomewhere else.> >> So, with Remus, the theory/goal is 0 downtime of your domU,whereas Pacemaker> >> simply minimizes downtime to a certain point - the time it takes todetect failure> >> and boot the new domU. > >> > > Actually, pacemaker can do live migrations with some limitations. Ifthe> > > > I have no experience with Remus, but from Nick''s description of it,it> > sounds like Remus might be a whole lot easier to set up and may be a > > good way to go if all you care about is failover for your domU''s, orif> > you really need instantaneous failover in the event of a servercrash.> > > All of this is well and good, but not to the point. > > I''m using pacemaker and DRBD. It works. it''s just a pain. > > Has anybody on this list actually used Remus in production? Is itready> for prime time? Or is it still beta (or alpha)? > > Miles > > -- > In theory, there is no difference between theory and practice. > In<fnord> practice, there is. .... Yogi Berra > > > > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xensource.com > http://lists.xensource.com/xen-users_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users The materials in this message are private and may contain Protected Healthcare Information. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Tuesday 07 September 2010 18:46:58 Miles Fidelman wrote:> Greg Woods wrote: > > On Tue, 2010-09-07 at 09:28 -0600, Nick Couchman wrote: > >> Pacemaker/DRBD, on the other hand, just synchronizes the data, > >> and if it detects that one of the domUs has died, starts it up somewhere > >> else. > >> > >> So, with Remus, the theory/goal is 0 downtime of your domU, whereas > >> Pacemaker > >> > >> simply minimizes downtime to a certain point - the time it takes to > >> detect failure > >> > >> and boot the new domU. > > > > Actually, pacemaker can do live migrations with some limitations. If the > > > > I have no experience with Remus, but from Nick''s description of it, it > > sounds like Remus might be a whole lot easier to set up and may be a > > good way to go if all you care about is failover for your domU''s, or if > > you really need instantaneous failover in the event of a server crash. > > All of this is well and good, but not to the point. > > I''m using pacemaker and DRBD. It works. it''s just a pain. > > Has anybody on this list actually used Remus in production? Is it ready > for prime time? Or is it still beta (or alpha)?Remus have very simply heartbeat(1s timeout), so in networks with heavy load it can failover without reason. To work properly it needs integrate with for example peacemaker(it''s not implemented), so pain will stay. Another problem is that Remus with disk replication doesn''t support failback. So in my opinion it''s not production ready. Regards -- Łukasz Oleś _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users