Hi,
I'm using "only-write-batch" and "read-batch" to
syncronize a debian root
filesystem with some personalized software on it, from a central server to
remote devices (all with RO filesystem, and perfectly identical among them).
In this way I'm able to push from my master device to an "updater"
server
and, from that, manage automatic updates on my remote devices.
But I have some performance issue:
- creating a batch image and save it on remote device is ok.
- when I want to apply my patch on a remote device, I have to wait 3/4
minutes to apply it, even if I changed just one small file.
with strace on rsync, I was able to see that the whole root filesystem is
passed in order to check checksum of all files, even if I don't have to
change them... why? This operation should be already done when I created
batch.
So is it possible to bypass the entire rootfs analysis (with chechsum, or
date and size)?
Is it possible using batch only to modify changed destination file without
considering the other ones in any way?
I tried with include/exclude options, but the whole rootfs is scanned the
same.
I also tried to search that on this forum, but without success.
I don't want/I can't use other software but rsync.
My rsync command is:
Thanks to who will reply me something.
--
Sent from: http://samba.2283325.n4.nabble.com/Samba-rsync-f2500462.html
