Hi, I'm pretty new to Gluster and I have a couple of questions. I would like to create a 2 host cluster over a fairly slow, sometimes unreliable network (http://www.torproject.org/). I'm not sure if what I want can be achieved using gluster. I'd like it to work as follows: Two servers, separated by the unreliable network mentioned above. A mount on each server which replicates all changes to the other server (two way) like a mirror. If the network is unresponsive at the time of a change, it waits until the network becomes responsive again and then sends the change. If I write to the local mount, I want it to be quick. I don't want to wait for the change to have been replicated over to the other server before the write completes. What sort of replication would I use to achieve this? Mike
http://www.gluster.org/docs/index.php/Understanding_AFR_Translator especially read the part about split brain. this is the situation you''ll have. also, using find to force auto-healing. there do seem to be problems with gluster 1.4pre5 when using AFR in the face of network interruptions. I''ve found that with my 2 AFR servers, if one is down for too long, gluster on the other server will just die--I presume it gets lonely and can''t go on. This is why I try not to date anyone whose co-dependant. I''d wait for the release of 1.4 since I believe this will be resolved. In the meantime you can try 1.3.12 which seem to work better, although not better at the same time. The AFR protocol in 1.4 is more efficient as is, I believe, it''s handling of split brain, but if you want to get cooking right now, start with 1.3.12 or wait a few weeks. Good luck. Keith At 01:24 AM 10/10/2008, Mike Cardwell wrote:>Hi, > >I''m pretty new to Gluster and I have a couple of questions. > >I would like to create a 2 host cluster over a fairly slow, sometimes >unreliable network (http://www.torproject.org/). I''m not sure if what I >want can be achieved using gluster. I''d like it to work as follows: > >Two servers, separated by the unreliable network mentioned above. A >mount on each server which replicates all changes to the other server >(two way) like a mirror. If the network is unresponsive at the time of a >change, it waits until the network becomes responsive again and then >sends the change. > >If I write to the local mount, I want it to be quick. I don''t want to >wait for the change to have been replicated over to the other server >before the write completes. > >What sort of replication would I use to achieve this? > >Mike > >_______________________________________________ >Gluster-users mailing list >Gluster-users at gluster.org >http://zresearch.com/cgi-bin/mailman/listinfo/gluster-users
http://www.gluster.org/docs/index.php/Understanding_AFR_Translator especially read the part about split brain. this is the situation you'll have. also, using find to force auto-healing. there do seem to be problems with gluster 1.4pre5 when using AFR in the face of network interruptions. I've found that with my 2 AFR servers, if one is down for too long, gluster on the other server will just die--I presume it gets lonely and can't go on. This is why I try not to date anyone whose co-dependant. I'd wait for the release of 1.4 since I believe this will be resolved. In the meantime you can try 1.3.12 which seem to work better, although not better at the same time. The AFR protocol in 1.4 is more efficient as is, I believe, it's handling of split brain, but if you want to get cooking right now, start with 1.3.12 or wait a few weeks. Good luck. Keith At 01:24 AM 10/10/2008, Mike Cardwell wrote:>Hi, > >I'm pretty new to Gluster and I have a couple of questions. > >I would like to create a 2 host cluster over a fairly slow, sometimes >unreliable network (http://www.torproject.org/). I'm not sure if what I >want can be achieved using gluster. I'd like it to work as follows: > >Two servers, separated by the unreliable network mentioned above. A >mount on each server which replicates all changes to the other server >(two way) like a mirror. If the network is unresponsive at the time of a >change, it waits until the network becomes responsive again and then >sends the change. > >If I write to the local mount, I want it to be quick. I don't want to >wait for the change to have been replicated over to the other server >before the write completes. > >What sort of replication would I use to achieve this? > >Mike > >_______________________________________________ >Gluster-users mailing list >Gluster-users at gluster.org >http://zresearch.com/cgi-bin/mailman/listinfo/gluster-users
http://www.gluster.org/docs/index.php/Understanding_AFR_Translator especially read the part about split brain. this is the situation you''ll have. also, using find to force auto-healing. there do seem to be problems with gluster 1.4pre5 when using AFR in the face of network interruptions. I''ve found that with my 2 AFR servers, if one is down for too long, gluster on the other server will just die--I presume it gets lonely and can''t go on. This is why I try not to date anyone whose co-dependant. I''d wait for the release of 1.4 since I believe this will be resolved. In the meantime you can try 1.3.12 which seem to work better, although not better at the same time. The AFR protocol in 1.4 is more efficient as is, I believe, it''s handling of split brain, but if you want to get cooking right now, start with 1.3.12 or wait a few weeks. Good luck. Keith At 01:24 AM 10/10/2008, Mike Cardwell wrote:>Hi, > >I''m pretty new to Gluster and I have a couple of questions. > >I would like to create a 2 host cluster over a fairly slow, sometimes >unreliable network (http://www.torproject.org/). I''m not sure if what I >want can be achieved using gluster. I''d like it to work as follows: > >Two servers, separated by the unreliable network mentioned above. A >mount on each server which replicates all changes to the other server >(two way) like a mirror. If the network is unresponsive at the time of a >change, it waits until the network becomes responsive again and then >sends the change. > >If I write to the local mount, I want it to be quick. I don''t want to >wait for the change to have been replicated over to the other server >before the write completes. > >What sort of replication would I use to achieve this? > >Mike > >_______________________________________________ >Gluster-users mailing list >Gluster-users at gluster.org >http://zresearch.com/cgi-bin/mailman/listinfo/gluster-users
http://www.gluster.org/docs/index.php/Understanding_AFR_Translator especially read the part about split brain. this is the situation you''ll have. also, using find to force auto-healing. there do seem to be problems with gluster 1.4pre5 when using AFR in the face of network interruptions. I''ve found that with my 2 AFR servers, if one is down for too long, gluster on the other server will just die--I presume it gets lonely and can''t go on. This is why I try not to date anyone whose co-dependant. I''d wait for the release of 1.4 since I believe this will be resolved. In the meantime you can try 1.3.12 which seem to work better, although not better at the same time. The AFR protocol in 1.4 is more efficient as is, I believe, it''s handling of split brain, but if you want to get cooking right now, start with 1.3.12 or wait a few weeks. Good luck. Keith At 01:24 AM 10/10/2008, Mike Cardwell wrote:>Hi, > >I''m pretty new to Gluster and I have a couple of questions. > >I would like to create a 2 host cluster over a fairly slow, sometimes >unreliable network (http://www.torproject.org/). I''m not sure if what I >want can be achieved using gluster. I''d like it to work as follows: > >Two servers, separated by the unreliable network mentioned above. A >mount on each server which replicates all changes to the other server >(two way) like a mirror. If the network is unresponsive at the time of a >change, it waits until the network becomes responsive again and then >sends the change. > >If I write to the local mount, I want it to be quick. I don''t want to >wait for the change to have been replicated over to the other server >before the write completes. > >What sort of replication would I use to achieve this? > >Mike > >_______________________________________________ >Gluster-users mailing list >Gluster-users at gluster.org >http://zresearch.com/cgi-bin/mailman/listinfo/gluster-users
Keith Freedman wrote:> http://www.gluster.org/docs/index.php/Understanding_AFR_Translator > > especially read the part about split brain. this is the situation > you'll have. > > also, using find to force auto-healing. > > there do seem to be problems with gluster 1.4pre5 when using AFR in the > face of network interruptions. I've found that with my 2 AFR servers, > if one is down for too long, gluster on the other server will just > die--I presume it gets lonely and can't go on. This is why I try not to > date anyone whose co-dependant. > > I'd wait for the release of 1.4 since I believe this will be resolved. > > In the meantime you can try 1.3.12 which seem to work better, although > not better at the same time. The AFR protocol in 1.4 is more efficient > as is, I believe, it's handling of split brain, but if you want to get > cooking right now, start with 1.3.12 or wait a few weeks.Thanks for the advice. I think I'll wait until 1.4 comes out and have a play then. Mike