| 1 |
Am 18.04.2011 10:12, schrieb Neil Bothwick: |
| 2 |
> On Mon, 18 Apr 2011 09:52:56 +0200, Florian Philipp wrote: |
| 3 |
> |
| 4 |
>>> it's tedious to install things |
| 5 |
>>> through an intermediary system all the time. The fullsize laptop, when |
| 6 |
>>> it gets its rebuild over the next week (it's been a windows 2k3 server |
| 7 |
>>> development system lately) |
| 8 |
> |
| 9 |
>> My strategy for getting Gentoo on a netbook with an SSD is to use NFS |
| 10 |
>> for PORTAGE_TMPDIR. Works nicely and makes less work than building |
| 11 |
>> everything remote. |
| 12 |
> |
| 13 |
> Doesn't using NFS slow compilation right down. I have a script on |
| 14 |
> the build host that enters the chroot and runs emerge -uD --changed-use |
| 15 |
> world, right after cron does emerge --sync, so the packages are |
| 16 |
> automatically available. Ass --usepkg to EMERGE_DEFAULT_OPTS on the |
| 17 |
> netbook and everything is transparent and no work at all (apart from a |
| 18 |
> couple of packages that won't build in the chroot). |
| 19 |
> |
| 20 |
> |
| 21 |
|
| 22 |
I haven't noticed any slowdown. I use a 100 MBit/s connection. That's |
| 23 |
nearly 12 MiB/s. The SSD has a write-speed of maybe 4-8 MiB/s. Actual |
| 24 |
throughput (monitored with iftop) was usually lower that 40 Mbit/s. |
| 25 |
Maybe latency was a bit higher and NFSv4 could have helped with that but |
| 26 |
I think it was negligible compared to the compiling performance of the |
| 27 |
Atom processor. |
| 28 |
|
| 29 |
Sure, a build host would have been better but it also meant more work. I |
| 30 |
also thought about using ATAoE, iSCSI or something alike to mount the |
| 31 |
SSD from a more powerful computer (using a live-system to avoid obvious |
| 32 |
problems when mounting the FS twice). Again - too much fuss. I usually |
| 33 |
just do security updates and then a full update every six months or so. |
| 34 |
|
| 35 |
Regards, |
| 36 |
Florian Philipp |
Archives