1 |
Luca Longinotti wrote: |
2 |
> The upgrade should actually be possible with PostgreSQL running, as the |
3 |
> blockers are there only to catch a little collision-protect problem (a |
4 |
> binary was moved from postgresql to libpq) and to warn about the pg-hier |
5 |
> USE flag disappearance. Pasting from a forums thread: |
6 |
> |
7 |
> "Well, the blocker was added for the simple reason of file-collisions. |
8 |
> With the new libpq/postgresql versions we've moved a file from the |
9 |
> postgresql package to the libpq package. Now, you have to emerge libpq |
10 |
> BEFORE postgresql, so when Portage was installing libpq it would bail |
11 |
> out with a "file collision found, blabla" error. The only way we have of |
12 |
> notifying the user of this is by adding a blocker. |
13 |
> The easiest and most correct way to upgrade would be this: |
14 |
> Code: |
15 |
> |
16 |
> FEATURES="-collision-protect" emerge --nodeps libpq |
17 |
> emerge postgresql |
18 |
> |
19 |
> The first one basically emerge libpq without checking for blockers and |
20 |
> disables the file-collision check, so that the new libpq will install |
21 |
> correctly, and then you can normally upgrade postgresql itself (which |
22 |
> you MUST do, both really *must* be upgraded!). " |
23 |
> |
24 |
> That should work and update your PostgreSQL install correctly. |
25 |
|
26 |
Yeah, I found that thread last night and followed the instructions. The |
27 |
emerge went fine, and postgres appeared to restart just fine... but when |
28 |
I tried to connect I discovered that it wasn't actually running! |
29 |
|
30 |
Well, long story short, I discovered a stale lock file in /tmp. The |
31 |
screwy thing was that the init.d script kept saying "ok" even though |
32 |
postgres wasn't actually running. That's a pretty bad bug, if you ask |
33 |
me, but I'm not a dev so... |
34 |
|
35 |
If anyone tries this, watch out for stale postgres lock files... in /tmp |
36 |
and starting with ".s.blah". I would have had only a few seconds of down |
37 |
time if I'd known that... and not much more than that if the init script |
38 |
had done it's job. |
39 |
|
40 |
b |
41 |
-- |
42 |
gentoo-server@g.o mailing list |