> My NFS client does not retry I/O. > Ganesha is reusing FD instead of opening file for every read request. > Sometimes it closes the fd even when there is a worker waiting still in > read() call on the same fd. As a result that worker returns read error and > ganesha drops the RPC. NFS client does not get any reply and it eventually times > out the I/O. > > I see 2 solutions. > > 1. NFS client implement I/O retry > 2. Ganesha open the file and try read again when file not opened error is > returned by FSAL. > 3. Ganesha open the file for every read RPC instead of reusing FD. > > NFS client is a legacy implementation and it is not going to be fixed. > Is there a way to achieve solutions 2 or 3? What version of Ganesha are you using? Which FSAL? Thanks Frank p.s. also consider subscribing to the new listserve

You can subscribe at https://lists.nfs-ganesha.org Are you using the support_ex functions? If not, I think there are some bugs. The non-support_ex functions are being deprecated (and in fact don’t exist in V2.6 or next). If you are using support_ex, your FSAL is responsible for protecting an in-use global FD, see how FSAL_VFS manages that. Without support_ex, the protection is supposed to be done with the mdcache content_lock, however, I’m not sure it’s always used. I think you can add your own protection. You could also yourself use a separate FD for each request. Although mdcache manages the lifetime of FDs, the FSAL really can do whatever it wants. Frank *From:*You Me [mailto:yourindianman@gmail.com] *Sent:* Monday, May 7, 2018 5:06 PM *To:* Frank Filz <ffilzlnx(a)mindspring.com&gt; *Cc:* NFS Ganesha Developers <nfs-ganesha-devel(a)lists.sourceforge.net>; devel(a)lists.nfs-ganesha.org *Subject:* Re: [Nfs-ganesha-devel] FD reuse in NFSv3 Ganesha 2.5.1 with my own FSAL. How do I subscribe to new listserve? --Satish On Mon, May 7, 2018 at 5:58 PM, Frank Filz <ffilzlnx(a)mindspring.com <mailto:ffilzlnx@mindspring.com>> wrote: > My NFS client does not retry I/O. > Ganesha is reusing FD instead of opening file for every read request. > Sometimes it closes the fd even when there is a worker waiting still in > read() call on the same fd. As a result that worker returns read error and > ganesha drops the RPC. NFS client does not get any reply and it eventually times > out the I/O. > > I see 2 solutions. > > 1. NFS client implement I/O retry > 2. Ganesha open the file and try read again when file not opened error is > returned by FSAL. > 3. Ganesha open the file for every read RPC instead of reusing FD. > > NFS client is a legacy implementation and it is not going to be fixed. > Is there a way to achieve solutions 2 or 3? What version of Ganesha are you using? Which FSAL? Thanks Frank p.s. also consider subscribing to the new listserve _______________________________________________ Devel mailing list -- devel(a)lists.nfs-ganesha.org To unsubscribe send an email to devel-leave(a)lists.nfs-ganesha.org