Spectrum

If I had to guess it would be the difference in recursion (-r) option between the two calls. If that is a typo and both calls are actually recursing, the difference could be added overhead going through the cygwin1.dll for cwrsync. I've never actually done any comparisons between rsync and cwrsync so it would be interesting to see if that is the difference. I have no clue about speed difference using rsync in daemon mode; hopefully someone else has some info on that.

This is not a DNS issue, it is a NETBIOS issue. Unless you are running your own DNS server locally, DNS has no bearing on name lookup for your local network. Samba is using NETBIOS over TCPIP for name resolution. Whenever your Windows box looks for http://tower it first checks it's hosts file. If no match is found it sends a broadcast to the local network asking if anyone is "named" tower. If no match is found it then queries a DNS server. I'm guessing android isn't configured to use NETBIOS over TCP/IP by default and I have no clue if it can be. It's Linux under the hood though, so you may be able to create/modify it's hosts file.

Yes, forcing the file_mode parameter allows for the files remain mapped as hidden/system on the original location; however, once you copy that file from the CIFS mount to another location, the copies files are going to have whatever perms you set using the file_mode parm. Mounting via CIFS and trying to maintain the Windows attributes is just not possible. You get either the 0777 that is default, or whatever you override it with. Example (Samba mapping HSA is on): Machine win1 (Windows OS) has a share foo and foo has the following files A SHR I file1 A H I file2 A S I file3 A I file4 A=Archive, S=System, H=Hidden, I=Not content indexed If \\win1\foo is mounted in an unRAID box using mount -t cifs //win1/foo /mnt/tmp doing an ls -l on /mnt/tmp would produce -rwxrwxrwx file1 -rwxrwxrwx file2 -rwxrwxrwx file3 -rwxrwxrwx file4 So if you copy any of those file to your array (the permissions would copy as listed above 0777), then the effective windows attributes would be A SH file1 A SH file2 A SH file3 A SH file4 If you instead mounted the share from an unRAID box with mount -t cifs //win1/foo /mnt/tmp -o file_mode=0644 doing an ls -l on /mnt/tmp would produce -rw-r--r-- file1 -rw-r--r-- file2 -rw-r--r-- file3 -rw-r--r-- file4 So if you copy any of those file to your array (the permissions would copy as listed above 0644), then the effective windows attributes would be file1 file2 file3 file4 So now from the above, if you copy those files to your array with the 777 perms then tried to view from a windows box, everything would be hidden and using the 644 perms, they would not. That is the case regardless of the original attributes on the windows box. Now as an example for tr0910's benefit: If you mount \\tower1\backup on win1 then copy the files across, they would maintain their windows attributes which would be mapped into *nix file perms. -r-x--x--x file1 -rw------x file2 -rw---x--- file3 -rw-r----- file4 You could then replicate from tower1 to tower2 using an NFS mount or using rsync as a daemon and you could preserve the permissions. Then you maintain the windows attribs through the whole chain and don't have to worry about things getting buggered due to CIFS mounts with the unRAID boxes. Now it's clear as mud right

Re-reading my response above it sounds a bit aggressive towards Brit's solution. That was not my intent in the least and apologies Brit if you read it and interpreted it that way. It's way too easy to mis-convey things when communicating via text Both are valid methods for solving your problem, however, there are repercussions if you remove Samba's ability to map hidden/system files (for Windows boxes) on your array. If you are fine with that than it is a good solution. At the end of the day, we just want things to work! If you want to avoid this problem all together, the easiest way would be to script your Windows machines to back up to Tower1 (rather than having Tower1 run the backup for each machine) and then use NFS mounts or rsync as a daemon process to backup from Tower1 to Tower2. All the perms would be maintained that way. There is cwRsync for Windows, or you could use xcopy or robocopy to script the backup from the Windows boxes to the Tower1 if you want to investigate that as an option. Glad Joe got you all fixed up on the reboot problem.

The downside to removing samba's mapping scheme is you lose the ability to have files marked as hidden/system on your array. This means if you copy a file across to the array that is a hidden/system file then copy it back off, it is no longer a hidden/system file. End of the world? Not really, but disabling the feature is using a shotgun instead of a scalpel. The only time this problem crops up is when mounting through cifs on the unRAID box. If you are copying files from one place in the array to another, the perms are preserved. If you are using NFS the perms are preserved. The CIFS driver mounts everything as 0777 which is a tad ridiculous IMO, but that's what it does. Solving the problem is using the file_mode option to override this behavior. Treating the symptom is blowing out Samba's ability to maintain hidden/system files for your Windows boxen.

Overide the perms with the mount command. mount -t cifs //server/share /mnt/tmp -o file_mode=0644 0644 mounts it as rw for root and read only for everyone else. That's the perms most of my files have in my array.

That is 100% normal. *nix systems don't show anything when you are entering a password at logon. Just type in your password and press enter.

Ouch $400/TB is steeeeap for rotational media even if they do look sweet. Reminds me of a rig I built back in the 90s. I took the guts of a Dell P2 system and mounted the MB, power supply, and 1 hard drive to a piece of plexi and hung it on the wall. It served up mp3s and acted as a web browser. Made for a good conversation piece

I haven't used a pch, but I would be a bit concerned if it is changing the file attributes of anything on the network and would start by fixing that problem. There's some info in this post about what is happening with Samba and Windows file attributes. You could check one of the affected iso's by executing ls -l /mnt/user/share/path/to/isofiles/*.iso from the command line on your unRaid box. for example ls -l /mnt/user/Movies/*.iso might give you something like -rwx------ 1 root root 13213880172 Aug 22 2010 15Minutes.iso* -rwx-----x 1 root root 17397927788 Jul 29 2010 2012.iso* -rwx------ 1 root root 2195218432 Oct 19 2008 A\ Clockwork\ Orange.iso* In that example 15Minutes and A Clockwork Orange would be un-hidden and 2012 would be hidden from a Windows machine's perspective. 2012 would be hidden because the world execute (that last x) is set. Before executing anything similar to what I've put below make sure that this is what is happening! The easiest way to remove this from all iso's in your array would be to execute find /mnt/user/ -name \*.iso -type f -exec chmod 700 {} \+ from a prompt on your unRaid box. To unhide all isos in your your movies share you could execute find /mnt/user/Movies -name \*.iso -type f -exec chmod 700 {} \+ If you have any questions or need help with syntax post back and I (or someone else) can help you out. Like I said earlier though, I would be more concerned if the PCH is changing files than this and would address that pronto.

http://lime-technology.com/forum/index.php?topic=10772.msg102770#msg102770

Troubleshooting unMenu would be a whole other topic and I'll leave that up to Joe L. You could always mount the shares on the Ubuntu box and run the md5 compare across the network. Not ideal, but not a horrible solution either and if this is just a one time thing it would get the job done.

Check out the md5deep package in unmenu. It does exactly what you are asking for.

What kind of drives are they? The short version is: Western Digital EARS with jumper on pins 7-8 you want to preclear without the -A option. Western Digital EARS without jumper on pins 7-8 you want to preclear with the -A option. Any other drive that is marked as Advanced Format (AF) you would want to preclear with the -A option. Any drive that is not marked as AF doesn't matter, either method will align on a boundary. If you want to change the drive you have already precleared (assuming it is still in a precleared state and has not had anything written to it or been added to an array) there is a way to do that quickly. Check in the preclear thread for more info. And thanks for reminding me the config tutorial needs to be updated

Keep it the way you have it. As is the only traffic that will go up to the router is traffic destined for the Internet, all system to system traffic will be handled by the switch. That said, unless the switch or the router is a real POS you probably wouldn't notice any real world difference. If you want to test raw bandwidth, you can use something like Iperf to get an idea what your network is capable of.

Depends on your definition of easy You have me curious though so if I come up with something I'll share the schematics!

I'd say nuke it from orbit but it looks like it might come after you

Weird, It doesn't look like a name resolution problem, it's resolving to the same IP address so it should be talking to the same box. What do you see if you hit it with a web browser and not just wget? The fact that it works without a password but not with one makes it sound like there is an authentication problem. Not sure the VPN config could mess that up, but anything is possible.

There are no firewalls of any kind in place so the server will respond to any requests that get through to it. Sounds more like a VPN or routing issue than an unRaid problem. Can you ping the server when connected to the VPN? Can you ping/connect to anything else on the network when connected to the VPN?

I doubt it would be worth the effort. 500Mbps is a little over 60MB/s so you wouldn't be seeing that much of a performance increase. Joe already said it's probably not your motherboard, but just for comparison I'm running unRaid on X7SPE-HF and getting ~800Mbps. Like Joe said, it's probably your OS. It's not all about raw bandwidth. I'm not sure if there would be a big difference in Vista/7, but XP will be slower when accessing remote file systems because it is using the older SMB filesystem rather than CIFS.

Ahh I missed that. Thanks for the heads up so I don't go redoing something that has already been done!

Hah I'm glad I gave all you guys something to play with over the weekend I see you got it working from command line, but you may need the JDK for Jperf to work. I know it works on Windows 7 x64 & x32 cause I've tried it on both; granted I do have the current JDK installed on all my boxen. Let me know if you install the JDK and it fixes it and I'll update the first post to help prevent others from falling down that hole. Cheater although surely someone around here has a 10gbps switch they can wow us with! As Benni mentioned 500Mbps is reasonable, but I would expect 750-800 for a single hop on a decent switch that isn't very busy

The more obvious tool to use would be md5deep to generate a list of md5 sums from the source, then use the list to check the destination. The list can also be saved for verifying integrity later if you want to hang on to it. I just didn't take the time to look for a Slackware package or to build one. I might build a Slack package an unMenu package for it some time but don't hold your breath

Quick and dirty would be diff -r /mnt/user/ntfs /mnt/disk3/tempo If you get no output from the diff command, everything is the same. Of course if you have written other data to the /mnt/disk3/tempo directory then this won't work quite as cleanly

Here is an example output on a Gbps link between my desktop and my unRaid box

I compiled Iperf 2.0.5 and got it up and running on unRAID to do some network analysis so I thought I would package it up and create an unMenu package for it. Iperf is a tool to test network bandwidth. It's very useful if you think you are having network issues, or if you just want to see what kind of transfer speeds are achievable between two machines on the network to which they are attached. It is a command line tool but there is a front end available, Jperf, that makes it easy to use and gives a nice pretty picture when it is finished with the test Jperf is distributed with an old version of the Windows Iperf which when used with version 2.0.5 will give incorrect results. I compiled Iperf 2.0.5 with Cygwin and made it and the cygwin1.dll available here. To use it with Jperf, download the binary and overwrite the iperf.exe distributed with Jperf with this one and the cygwin1.dll. It is a fully functional binary so you can use it from the command line as well if you prefer. If you want to test from a Linux machine, most distros have a package available for Iperf, just make sure you are getting the 2.0 release not 1.0. Mac users, you're on your own; but I bet you could compile it from source if there is not a package of some sort available! Mac instructions from Benni-chan: You have to have xcode installed, since you need to compile it from the source. Download and extract the source and use following commands ./configure make sudo make install Note that installing Iperf on unRaid just installs the binary files, it doesn't actively start the process which is not something you want running all the time anyway. You need to log into your unRAID box and start Iperf when you want to run a test and stop it with CTRL-C when you are finished. There is a good tutorial on using Iperf here. I don't have any affiliation with that site, but it's the first one that Google gave up when I searched for Iperf tutorial To keep the GPL/BSD police happy I have made the source code available for Iperf and Cygwin but I'd prefer you use the project's bandwidth instead of mine If you have any problems, questions, or suggestions let me know.