Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Nick, The 2 URLs (Which in fact are just one same article) are interesting. However, the setup that this guy describes looks fairly simple, and mostly "home lab" oriented to what I understand. In my personal case, that's something to put in production that I need, and production needs something proven. After some months investigating the DRBD stuff, I now have the strong feeling that DRBD IS proven. First by looking at the huge number of setups all over the world. If the product was bad, at that time one would know. Second by the fact that many industrialized solutions have been implemented, some being certified by various actors. It wouldn't have that success if had strong lacks or drawbacks. I didn't know of that Glusterfs product. Glad to know that it exists, but I guess that if the objective is real production, I guess it would be a wise attitude to let it "dry" a bit, and see if it gains more success in the future. You can't trust a product solely because its features look great, right ? Confirming that feeling, if you go by the end of the URL, in the comments area, there's an interesting mention of a guy who obviously belongs to the development team, and who says more or less that iSCSI is not the right technology to adopt for large scale Glusterfs deployments, it has side effects that can lead to data corruption (He doesn't explicitly say so, but the effects he describes clearly indicate it might certainly!). Then, the IET that guy makes use of has some known lacks, in terms of performance as well as functionality, especially in the virtualization area (I faced a couple of problems with round-robin accesses under VMware for instance). I'm surprised that he doesn't know them and don't propose something more serious like SCST. If then you chose NFS to export your datas, it's very likely that you will lose important features such as VMware Thin Provisionning, or should I say, you will use it but it won't save any space, too bad... This makes me feel like this guy, as honest as his enthusiasm might be (And I believe it is), mainly relies on the announced features rather than on a real experience of the product itself. And that's what makes me prudent. Talking about production, I mean. Also, my understanding was that ZFS was under FreeBSD licensing and as far as I know there was nothing available for our common GNU distributions, I'm surprised to see that it has changed, may be the Oracle effect. If so, that might be of interest, but in that case nothing prevents you from using ZFS with DRBD, right ? At least that's my understanding... So to summarize, it's interesting to know that this product exists, but it's urgent to wait before implementing it. There are tons of DRBD clusters experience available everywhere, an official support, you have very few for Glusterfs to what I can see. You have tons of references implementing DRBD and Pacemaker. How many companies do you know implementing this Glusterfs ? In my case, none. And you probably have very few answers to your question " Any feedback regarding to IET+GlusterFS or ZFS, or other setups", simply because nobody or so can really tell... DRBD and Pacemaker stuffs are effectively a bit complex, but clusterizing data access IS a complex (And sensible!) task by nature, and you get nothing for nothing. I think that the best advice I may give you is spend the time it requires to master these products, you won't regret it! I prefer suffering a bit and trusting my datas rather than trusting something that looks so simple and smart, and then having to explain my customers that their datas are definitely gone... 20 years of IT made a prudent guy of me! :) But if your goal is not production and only a test platform or whatever simple not needing specific performance level and tortured scenarios, then why not going for Glusterfs... In the end, it's mainly a matter of use case... And unfortunately, no, at the moment I don't know of any other serious data clustering offerings, except from array manufacturers or dedicated actors such as Datacore or Falconstor, but it costs an eye and I personally had the opportunity to see big bad situations with such products, with official support guys not being able to explain why it had sucked up and not being able to tell when it would go back to production. VMware has introduced a - paying - Virtual Storage Appliance as of vSphere 5.0, which proposes a cluster of replicated data volumes, but honestly it doesn't compare to the DRBD/Pacemaker features, many lacks, many limitations. May be it will get better in the future, but at the moment it's not competitive. I'm a pro-VMware, I should say it's good, but objectively I can't! So that's why I have been so happy to discover DRBD clusters last year and why I spend so much time those last few months to understand how they work and try to master them. Not that easy you're right, but I know I won't regret it! Best regards, Pascal. -----Message d'origine----- De : drbd-user-bounces at lists.linbit.com [mailto:drbd-user-bounces at lists.linbit.com] De la part de Nick Khamis Envoyé : dimanche 30 octobre 2011 02:05 À : linux-ha at lists.linux-ha.org; drbd-user at lists.linbit.com Objet : [DRBD-user] DRBD Alternatives (Apples vs. Apples) As I look to find a other native file replication solutions (i.e., network raid 1), I found: http://t.co/jeo6QoH5, and http://www.bauer-power.net/2011/08/roll-your-own-fail-over-san-cluster.html With that in mind, has anyone else here looked into alternatives? An apples with applies comparison, so no talk about SANs+LUN. Any feedback regarding to IET+GlusterFS or ZFS, or other setups is greatly appreciated? Looking forward to getting your feedback, Nick. _______________________________________________ drbd-user mailing list drbd-user at lists.linbit.com http://lists.linbit.com/mailman/listinfo/drbd-user