FWIW: Running a session off an NFS share

[Milian Wolff]

On Freitag, 18. Januar 2019 13:00:26 CET René J.V. Bertin wrote:
> Hi,
> 
> I sometimes open projects that are on an NFS share (because running KDevelop
> over remote X11 tends to be problematic). NFS works, but can be
> cruciatingly slow even if only the project sources are on the server (which
> is on the same LAN where I routinely get around 8Mbps file transfer
> speeds).
> 
> As an experiment I just ran KDevelop through `eatmydata` (slightly patched
> to show stats) on a project on an NFS share and that made the experience a
> *lot* better. After quitting KDevelop I got this confirmation why:
> 
> kdevelop: eatmydata swallowed 106710 time(s)
> 
> That means the relatively short session in which I only made a few minor
> changes issued over 100000 explicit sync calls and/or operations supposed
> to be done in synchronous mode (O_SYNC etc).
> 
> "Eating" those syncs didn't cause any problems (possibly because my patch
> also does an explicit sync at exit) but note that I was also logged to the
> server (a Mac) to do the builds locally.
> 
> My patched eatmydata hack: github:RJVB/eatmydata .

Please find out where we are syncing, e.g. via hotspot's off-CPU profiling 
capabilities. I would suspect it's mostly when we sync config files?

Note that not syncing those after a change is often a very bad idea, as the 
user may then lose the settings after a crash. And sadly, KDevelop can still 
crash often...

-- 
Milian Wolff
mail at milianw.de
http://milianw.de
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: This is a digitally signed message part.
URL: <http://mail.kde.org/pipermail/kdevelop-devel/attachments/20190127/d3d32828/attachment.sig>