----- Original Message ----- > From: "Emmanuel Dreyfus" <[email protected]> > To: "Raghavendra Gowdappa" <[email protected]>, "Manikandan Selvaganesh" > <[email protected]> > Cc: "Vijaikumar Mallikarjuna" <[email protected]>, "Raghavendra Talur" > <[email protected]>, "Gluster Devel" > <[email protected]> > Sent: Wednesday, January 6, 2016 1:21:19 AM > Subject: Re: [Gluster-devel] quota.t hangs on NetBSD machines > > Raghavendra Gowdappa <[email protected]> wrote: > > > I figured out this issue occurs when "trickling-writes" is on in > > write-behind > > Do you have a idea of what fails at system level? Is it a portability > bug, or something else?
No, its not a portability issue. It should have occurred at similar probability on Linux machines too. Does Linux issue frequent fsyncs compared to NetBsd? The only thing that could've made a difference is an fsync from kernel to Glusterfs. If there is an fsync after hung write, that would've unwound the write call. Application cannot do it as it is hung in write call. I am wondering whether kernel could've triggered an fsync. But, I myself is not convinced by this explanation. Lets wait till the test results are available so that at least we are sure of RCA. > > -- > Emmanuel Dreyfus > http://hcpnet.free.fr/pubz > [email protected] > _______________________________________________ Gluster-devel mailing list [email protected] http://www.gluster.org/mailman/listinfo/gluster-devel
