Here's a better patch. It fixes the following problems: * nbd-client.c in the last patch had no GPL info * nbd-client.c's opennet() in the the last patch returned positive numbered errors. * nbd-server.c in 2.7.3 called wait() with a NULL pointer, causing nbd-server to hang in wait() when SIGCHLD was delivered and more than one client had connected. It still doesn't fix the issue with the NBD not being noticed by the software RAID as being reconnected, I suspect that will be a kernel-level fix. I'm working only on the user-land tools right now. On Thu, 28 Apr 2005 16:13:00 -0500, J. Ryan Earl wrote > I'm not sure, it's somewhere in there at the block level, I'm sure some > kernel guru would know. I wonder how the SCSI layer does it... > > -ryan > > -----Original Message----- > From: nbd-general-admin@lists.sourceforge.net > [mailto:nbd-general-admin@lists.sourceforge.net]On Behalf Of Roy > Keene Sent: Thursday, April 28, 2005 3:30 PM To: J. Ryan Earl Cc: > nbd-general@lists.sourceforge.net Subject: RE: [Nbd] ndb-client > 2.7.3 patch > > As far as I can tell it doesn't sense that nbdX comes back to life. > But it keeps you from having to restart the "nbd-client" when your client > comes back, bringing us a step closer. > > I haven't figured out how to manually tell it to start the resync, > any clues ? > > On Thu, 28 Apr 2005, J. Ryan Earl wrote: > > > Interesting, this is exactly what I use NBD for (RAID1 over the network). > > Does this allow a resync to happen automatically when the nbd-client > > restores its connection to the nbd-server? ie Will md sense when the nbd > is > > available again? > > > > -ryan > > > > -----Original Message----- > > From: nbd-general-admin@lists.sourceforge.net > > [mailto:nbd-general-admin@lists.sourceforge.net]On Behalf Of Roy Keene > > Sent: Thursday, April 28, 2005 3:05 PM > > To: nbd-general@lists.sourceforge.net > > Subject: [Nbd] ndb-client 2.7.3 patch > > > > > > I've partially re-written the nbd-client from nbd-2.7.3 (sorry it's not > > against the 2.8.x or CVS, I needed this working quickly) to be more of a > > daemon and not exit if the connection goes away. > > > > I use this for my software RAID of network block devices so I want it to > > reconnect once the host becomes available again without intervention. > > > > Feel free to license it however you wish. > > > > Most changes only affect nbd-client.c, but I did update cliserv.h to > > include "err_noexit()" that behaves the same as err() but does not call > > exit(). > > > > ------------------------------------------------------- > SF.Net email is sponsored by: Tell us your software development > plans! Take this survey and enter to win a one-year sub to SourceForge.net > Plus IDC's 2005 look-ahead and a copy of this survey > Click here to start! http://www.idcswdc.com/cgi-bin/survey?id=105hix > _______________________________________________ > Nbd-general mailing list > Nbd-general@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/nbd-general > > ------------------------------------------------------- > SF.Net email is sponsored by: Tell us your software development > plans! Take this survey and enter to win a one-year sub to SourceForge.net > Plus IDC's 2005 look-ahead and a copy of this survey > Click here to start! http://www.idcswdc.com/cgi-bin/survey?id=105hix > _______________________________________________ > Nbd-general mailing list > Nbd-general@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/nbd-general -- Roy Keene rkeene@...51...
Attachment:
nbd-client-daemon-2.diff
Description: Binary data