| 1 |
On 9/8/06, Vladimir G. Ivanovic <vgivanovic@×××××××.net> wrote: |
| 2 |
> On Thu, 2006-09-07 at 15:30 -0400, Andy Dustman wrote: |
| 3 |
> > gentoo-server is probably not the best list for this discussion, but I |
| 4 |
> > suspect many or most of the people on it will have an interest in |
| 5 |
> > this, so it is somewhat-on-topic. If there's any interest/traction, |
| 6 |
> > then I'll redo this as a GLEP. If you like the idea but think it |
| 7 |
> > should be discussed off this list, contact me directly. |
| 8 |
> > |
| 9 |
> > There are two main issues with the Portage tree as I see it: |
| 10 |
> > |
| 11 |
> > 1) There's no way to get just security updates. |
| 12 |
> |
| 13 |
> I run this script every hour: |
| 14 |
> |
| 15 |
> #!/bin/sh |
| 16 |
> glsa-check -f new 2>/dev/null |
| 17 |
> [[ $? -eq 0 ]] || echo "glsa-check: error" |
| 18 |
> |
| 19 |
> Does this not do what you're looking for? |
| 20 |
|
| 21 |
No, but I do something like this already in my own update/sync script, |
| 22 |
and the output is mailed to me: |
| 23 |
|
| 24 |
emerge --sync --quiet --nocolor |
| 25 |
glsa-check --dump affected |
| 26 |
emerge world --pretend --update --changelog --columns --nocolor |
| 27 |
|
| 28 |
Getting security-only updates is not a major issue for me, but some |
| 29 |
people really want them for long-term stability. |
| 30 |
|
| 31 |
> > 2) Even with recent metadata cache update improvements, it still takes |
| 32 |
> > a long time, and a lot of resources, to sync the tree. |
| 33 |
> |
| 34 |
> I do my updates early in the morning when I'm asleep, so I don't really |
| 35 |
> care how long they take (as long as it's less than 4 hours). |
| 36 |
> |
| 37 |
> Why does update speed matter to you? |
| 38 |
|
| 39 |
Speed is not the only, or even primary, issue, though it's a little |
| 40 |
slow (approximately 10 minutes, mostly disk-bound). I do not have a |
| 41 |
lack of network bandwidth (225 Mbps backbone, and gigabit throughout |
| 42 |
most of the local network). Disk bandwidth and load on the mirrors |
| 43 |
(network, disk, CPU) are probably more important. You can throw |
| 44 |
hardware at the problem (i.e. add more mirrors), though as I |
| 45 |
understand it, this depends on the goodwill of volunteers. |
| 46 |
|
| 47 |
Right now, there are about 150K files in the Portage tree; this |
| 48 |
includes directories as they are sent in the file list by rsync.That's |
| 49 |
150K stat() calls (at least) on both the mirror and the client. Having |
| 50 |
a seperate updates overlay should reduce this dramatically; |
| 51 |
immediately after a new release, the updates overlay would be empty. |
| 52 |
The question is, How quickly would an updates overlay grow? |
| 53 |
|
| 54 |
Mike Kelly mentioned emerge-delta-webrsync, so I took a look at the |
| 55 |
size of the entire compressed Portage tree (31 M) vs. the size of the |
| 56 |
compressd patches (250 K, rough average by eye), and this looks like |
| 57 |
approximately 1% per day. If that's really the case, then an |
| 58 |
updates-only overlay may not be viable. However, getting compressed |
| 59 |
patches is probably a big improvement over rsync, so I'll have to try |
| 60 |
it out. |
| 61 |
-- |
| 62 |
This message has been scanned for memes and |
| 63 |
dangerous content by MindScanner, and is |
| 64 |
believed to be unclean. |
| 65 |
-- |
| 66 |
gentoo-server@g.o mailing list |
Archives