sorry, accidentally hit reply instead of reply all
On Fri, Jan 14, 2011 at 10:05 AM, Gareth Evans <agr...@gmail.com> wrote: > Okay I guess the recursive thing makes sense, I hope IIS establishes > watches on the specific directories it's interested in. > The tree could indeed be very deep. > I guess the only way to know is to test with it once I can repeat the same > operations as you and get the same output. > > I don't know what i'm doing differently, but it certainly didn't detect the > move/copy/modify operations when I tested it. > Would the local server filesystem make a difference? I doubt it, but worth > asking > > There has to be something different between our two tests, there can't be > anything I'm doing wrong specifically that causes it to not work, it's a > pretty simple test. > I was definitely making the changes in the directory that I had the > notify/watch established for > > Do you see the 'client only requested 32 bytes marshalling 58 (?) bytes' > message in your log output? I always saw that error when I was testing and > it didn't work. > From reading the source, that error is generated immediately before a > return false from the notify response. > How long is your UNC path? Mine was a reasonable length: > That's 52 bytes by itself, plus a filename could be where that ~58 bytes > number came from > \\smbtest\mwh-webstore\clientname\site.co.nz\htdocs > > I think that if I triggered the changes from a samba share that it was > handled correctly, but the changes on the shell didn't. > I'd need to test this specific scenario to confirm if this was the case. > > In our scenario, our FTP servers are modifying files via a mounted NFS > share, or things are changed directly on the console (e.g. unzipping an > archive of site files) > > I'm at the office right now, and my test ubuntu server with 3.5.6 is on it > is at home - I can go there if you need more testing info or want me to try > specific scenarios. > > > Gareth > > > On Fri, Jan 14, 2011 at 9:55 AM, Volker Lendecke < > volker.lende...@sernet.de> wrote: > >> On Fri, Jan 14, 2011 at 09:39:13AM +1300, Gareth Evans wrote: >> > Looks like you're using a mapped drive letter rather than a UNC path >> -that >> > could be contributing. >> >> Yes. Just tried with a UNC path. Same result. >> >> > It sounds like what you're doing to test is fairly equivilent to what I >> was >> > doing but in this case, i'm seeing what I would have expected to see in >> my >> > testing. >> > >> > What OS is your client machine? XP? Win7? >> >> Win7 for this test. >> >> > If a smb client requests a recursive watch, such as IIS would (because >> it >> > would register the notify on the root folder and expect to be told if >> > something in bin subfolder changes) does samba register notifications in >> a >> > recursive way (i.e. subfolders) >> >> It does internally. This means that we get recursive change >> notifies that are triggered via other CIFS clients. >> >> For example the >> >> File: z:\mput\file00000168.exe Deleted >> >> from my last message was triggered by smbclient connected to >> the same share. Doing a "rm /data/tmp/mput/xxx" on the Linux >> server shell would not have triggered it. >> >> We do not get recursive changes triggered by local Unix >> clients. Given the non-recursive nature of the inotify API, >> his would be potentially prohibitively expensive. We might >> have to walk a directory tree arbitrarily large. >> >> With best regards, >> >> Volker Lendecke >> >> -- >> SerNet GmbH, Bahnhofsallee 1b, 37081 Göttingen >> phone: +49-551-370000-0, fax: +49-551-370000-9 >> AG Göttingen, HRB 2816, GF: Dr. Johannes Loxen >> > > -- To unsubscribe from this list go to the following URL and read the instructions: https://lists.samba.org/mailman/options/samba