Hi Gordan, Thanks for your feedback on unfsd. We will have a look. How's the experience on 3.0.2rc1, performance wise, between a Gluster client and server using the native protocol ? Did you see that worsen in comparison to 2.0.9 or improve compared to 2.0.9 ? Would be very interested to know. Regards, Tejas. ----- Original Message ----- From: "Gordan Bobic" <gordan@xxxxxxxxxx> To: "Gluster List" <gluster-devel@xxxxxxxxxx> Sent: Thursday, February 4, 2010 11:58:39 PM GMT +05:30 Chennai, Kolkata, Mumbai, New Delhi Subject: Re: gluster 3.0.0 catastrophic crash during basic file creation test Tejas N. Bhise wrote: > Besides that, if you have recently upgraded to 3.0.0, > please consider 3.0.2 which would be out very soon > ( your can even try 3.0.2rc1 ). It has much better > performance that previous versions too. I'm no longer that convinced. My observation is that latencies in the glfs+unfsd use-case are _much_ higher with 3.0.2rc1 than they were with 2.0.9. So much higher that the setup is nearly unusable for /home. I haven't observed this slow-down with other things on glfs root, though, only with unfsd exports. io-cache and write-behind translators seem to help a lot, but I'm still weary of using any performance translators at the moment, especially with the spurious disconnects I mentioned in a separate thread. Gordan _______________________________________________ Gluster-devel mailing list Gluster-devel@xxxxxxxxxx http://lists.nongnu.org/mailman/listinfo/gluster-devel
