From jajcus at jajcus.net Sun Mar 9 16:51:21 2014 From: jajcus at jajcus.net (Jacek Konieczny) Date: Sun, 09 Mar 2014 16:51:21 +0100 Subject: testing needed: systemd unit files for PostgreSQL Message-ID: <531C8DF9.5000106@jajcus.net> Hi, I found the init script for our postgresql a bit unreliable when used under systemd. Additionally, most systemd process monitoring features are lost when managing a service through the legacy init scripts. I decided to add native systemd units for PostgreSQL, but that was not trivial, as our init script supports multiple PostgreSQL instances, in arbitrary locations (defined by $PG_DB_CLUSTERS in /etc/sysconfig/postgresql). Ideally, 'service postgresql start/stop/reload' should work as it used to ? start/stop/reload all the instances. But, at the same time, we want every instance being separately managed by systemd. I have implemented this in postgresql-9.3.3 by making 'postgresql.service' a dummy service, which just pulls 'postgresql.target', which Requires each individual instance defined through 'postgresql at .service'. That is how it works (with two clusters configured: /var/lib/pgsql and /var/lib/pgsql-1): Starting: > [root at lolek ~]# service postgresql start > Redirecting to /bin/systemctl --output=cat start postgresql.service > [root at lolek ~]# service postgresql status > Redirecting to /bin/systemctl --output=cat status postgresql.service > postgresql.service - PostgreSQL server > Loaded: loaded (/lib/systemd/system/postgresql.service; enabled) > Active: active (exited) since Sun 2014-03-09 16:45:22 CET; 4s ago > Process: 11825 ExecStart=/bin/systemctl list-dependencies postgresql.target (code=exited, status=0/SUCCESS) > Process: 11823 ExecStart=/bin/systemctl is-active -q postgresql.target (code=exited, status=0/SUCCESS) > Main PID: 11825 (code=exited, status=0/SUCCESS) > > Starting PostgreSQL server... > Started PostgreSQL server. > [root at lolek ~]# service postgresql at var-lib-pgsql status > Redirecting to /bin/systemctl --output=cat status postgresql at var-lib-pgsql.service > postgresql at var-lib-pgsql.service - PostgreSQL database server at /var/lib/pgsql > Loaded: loaded (/lib/systemd/system/postgresql at .service; enabled) > Active: active (running) since Sun 2014-03-09 16:45:22 CET; 15s ago > Process: 11757 ExecStop=/usr/bin/pg_ctl -D %f stop -w -s -m fast (code=exited, status=0/SUCCESS) > Process: 11799 ExecStart=/usr/bin/pg_ctl -D %f start -w -s (code=exited, status=0/SUCCESS) > Main PID: 11807 (postgres) > CGroup: /system.slice/system-postgresql.slice/postgresql at var-lib-pgsql.service > ??11807 /usr/bin/postgres -D /var/lib/pgsql > ??11811 postgres: checkpointer process > ??11812 postgres: writer process > ??11813 postgres: wal writer process > ??11814 postgres: autovacuum launcher process > ??11815 postgres: stats collector process > > [1-1] LOG: ending log output to stderr > [1-2] HINT: Future log output will go to log destination "syslog". > LOG: ending log output to stderr > HINT: Future log output will go to log destination "syslog". > [2-1] LOG: database system was shut down at 2014-03-09 15:44:19 GMT > [2-1] LOG: database system is ready to accept connections > [2-1] LOG: autovacuum launcher started > [3-1] LOG: connection received: host=[local] > [4-1] LOG: connection authorized: user=postgres database=postgres > Started PostgreSQL database server at /var/lib/pgsql. > [root at lolek ~]# service postgresql at var-lib-pgsql\\x2d1 status > Redirecting to /bin/systemctl --output=cat status postgresql at var-lib-pgsql-1.service > postgresql at var-lib-pgsql\x2d1.service - PostgreSQL database server at /var/lib/pgsql-1 > Loaded: loaded (/lib/systemd/system/postgresql at .service; enabled) > Active: active (running) since Sun 2014-03-09 16:45:22 CET; 27s ago > Process: 11756 ExecStop=/usr/bin/pg_ctl -D %f stop -w -s -m fast (code=exited, status=0/SUCCESS) > Process: 11800 ExecStart=/usr/bin/pg_ctl -D %f start -w -s (code=exited, status=0/SUCCESS) > Main PID: 11808 (postgres) > CGroup: /system.slice/system-postgresql.slice/postgresql at var-lib-pgsql\x2d1.service > ??11808 /usr/bin/postgres -D /var/lib/pgsql-1 > ??11816 postgres: checkpointer process > ??11817 postgres: writer process > ??11818 postgres: wal writer process > ??11819 postgres: autovacuum launcher process > ??11820 postgres: stats collector process > > [1-1] LOG: ending log output to stderr > [1-2] HINT: Future log output will go to log destination "syslog". > [2-1] LOG: database system was shut down at 2014-03-09 16:44:19 CET > LOG: ending log output to stderr > HINT: Future log output will go to log destination "syslog". > [2-1] LOG: database system is ready to accept connections > [2-1] LOG: autovacuum launcher started > [3-1] LOG: connection received: host=[local] > [4-1] LOG: connection authorized: user=postgres database=postgres > Started PostgreSQL database server at /var/lib/pgsql-1. Reloading: > [root at lolek ~]# service postgresql reload > Redirecting to /bin/systemctl --output=cat reload postgresql.service > [root at lolek ~]# service postgresql at var-lib-pgsql status > Redirecting to /bin/systemctl --output=cat status postgresql at var-lib-pgsql.service > postgresql at var-lib-pgsql.service - PostgreSQL database server at /var/lib/pgsql > Loaded: loaded (/lib/systemd/system/postgresql at .service; enabled) > Active: active (running) since Sun 2014-03-09 16:45:22 CET; 44s ago > Process: 11757 ExecStop=/usr/bin/pg_ctl -D %f stop -w -s -m fast (code=exited, status=0/SUCCESS) > Process: 11869 ExecReload=/usr/bin/pg_ctl -D %f reload -w -s (code=exited, status=0/SUCCESS) > Process: 11799 ExecStart=/usr/bin/pg_ctl -D %f start -w -s (code=exited, status=0/SUCCESS) > Main PID: 11807 (postgres) > CGroup: /system.slice/system-postgresql.slice/postgresql at var-lib-pgsql.service > ??11807 /usr/bin/postgres -D /var/lib/pgsql > ??11811 postgres: checkpointer process > ??11812 postgres: writer process > ??11813 postgres: wal writer process > ??11814 postgres: autovacuum launcher process > ??11815 postgres: stats collector process > > HINT: Future log output will go to log destination "syslog". > [2-1] LOG: database system was shut down at 2014-03-09 15:44:19 GMT > [2-1] LOG: database system is ready to accept connections > [2-1] LOG: autovacuum launcher started > [3-1] LOG: connection received: host=[local] > [4-1] LOG: connection authorized: user=postgres database=postgres > Started PostgreSQL database server at /var/lib/pgsql. > Reloading PostgreSQL database server at /var/lib/pgsql. > [3-1] LOG: received SIGHUP, reloading configuration files > Reloaded PostgreSQL database server at /var/lib/pgsql. Stopping: > [root at lolek ~]# service postgresql stop > Redirecting to /bin/systemctl --output=cat stop postgresql.service > [root at lolek ~]# service postgresql at var-lib-pgsql status > Redirecting to /bin/systemctl --output=cat status postgresql at var-lib-pgsql.service > postgresql at var-lib-pgsql.service - PostgreSQL database server at /var/lib/pgsql > Loaded: loaded (/lib/systemd/system/postgresql at .service; enabled) > Active: inactive (dead) since Sun 2014-03-09 16:46:11 CET; 1s ago > Process: 11891 ExecStop=/usr/bin/pg_ctl -D %f stop -w -s -m fast (code=exited, status=0/SUCCESS) > Process: 11869 ExecReload=/usr/bin/pg_ctl -D %f reload -w -s (code=exited, status=0/SUCCESS) > Process: 11799 ExecStart=/usr/bin/pg_ctl -D %f start -w -s (code=exited, status=0/SUCCESS) > Main PID: 11807 (code=exited, status=0/SUCCESS) > > Reloading PostgreSQL database server at /var/lib/pgsql. > [3-1] LOG: received SIGHUP, reloading configuration files > Reloaded PostgreSQL database server at /var/lib/pgsql. > Stopping PostgreSQL database server at /var/lib/pgsql... > [4-1] LOG: received fast shutdown request > [5-1] LOG: aborting any active transactions > [3-1] LOG: autovacuum launcher shutting down > [2-1] LOG: shutting down > [3-1] LOG: database system is shut down > Stopped PostgreSQL database server at /var/lib/pgsql. > [root at lolek ~]# service postgresql at var-lib-pgsql\\x2d1 status > Redirecting to /bin/systemctl --output=cat status postgresql at var-lib-pgsql-1.service > postgresql at var-lib-pgsql\x2d1.service - PostgreSQL database server at /var/lib/pgsql-1 > Loaded: loaded (/lib/systemd/system/postgresql at .service; enabled) > Active: inactive (dead) since Sun 2014-03-09 16:46:11 CET; 8s ago > Process: 11890 ExecStop=/usr/bin/pg_ctl -D %f stop -w -s -m fast (code=exited, status=0/SUCCESS) > Process: 11868 ExecReload=/usr/bin/pg_ctl -D %f reload -w -s (code=exited, status=0/SUCCESS) > Process: 11800 ExecStart=/usr/bin/pg_ctl -D %f start -w -s (code=exited, status=0/SUCCESS) > Main PID: 11808 (code=exited, status=0/SUCCESS) > > Started PostgreSQL database server at /var/lib/pgsql-1. > Reloading PostgreSQL database server at /var/lib/pgsql-1. > [3-1] LOG: received SIGHUP, reloading configuration files > Reloaded PostgreSQL database server at /var/lib/pgsql-1. > Stopping PostgreSQL database server at /var/lib/pgsql-1... > [4-1] LOG: received fast shutdown request > [5-1] LOG: aborting any active transactions > [2-1] LOG: shutting down > [3-1] LOG: database system is shut down > Stopped PostgreSQL database server at /var/lib/pgsql-1. The problem is 'service postgresql status' gives little useful information now (more information can be obtained when querying individual instances), but except that I think the configuration works quite well. One can also disable the 'postgresql.service' and manually manage the 'postgresql at something.service' units. Please, check it out and tell if something is broken. Greets, Jacek From draenog at pld-linux.org Thu Mar 20 20:03:12 2014 From: draenog at pld-linux.org (Kacper Kornet) Date: Thu, 20 Mar 2014 20:03:12 +0100 Subject: poldek and rpm: order of packages during removal Message-ID: <20140320190312.GA28484@camk.edu.pl> Short version: I would like to remove form poldek --noorder option passed to rpm by poldek during package removal. Long version: Currently the order of packages during removal by poldek is broken in Th. For example: $ rpm -q --qf '[%{REQUIRENAME} %{REQUIREFLAGS:deptype}\n]' test-sub /bin/sh preun rpmlib(PayloadIsLzma) rpmlib $ rpm -q --qf '[%{REQUIRENAME} %{REQUIREFLAGS:deptype}\n]' test-sub2 /bin/sh preun test-sub preun rpmlib(PayloadIsLzma) rpmlib So test-sub2 should be removed before test-sub. However: $ poldek -evv test-sub test-sub2 Setting temporary cache directory path to /root/tmp/poldek-cache-root Setting cache directory path to /root/tmp/poldek-cache-root Enabling multilib mode, transaction color = 3 Trying test-sub Trying test-sub2 mark test-sub-1-0.1.x86_64 mark test-sub2-1-0.1.x86_64 Processing dependencies... Preparing package set... There are 2 packages to remove: R test-sub-1-0.1.x86_64 test-sub2-1-0.1.x86_64 Proceed? [N/y] y Running etckeeper.sh --erase -v --root / --noorder test-sub-1-0.1.x86_64 test-sub2-1-0.1.x86_64... sub script sub2 script Poldek orders packages assuming that rpm starts to remove them from the last one. However it is true only when --noorder flag is *not* used and packages don't depend on each other. With flag --noorder rpm in Th seems to proceed packages starting from the first one. At the same time current rpm in Th seems to be able to order packages for removal by himself. $ rpm -e test-sub2 test-sub sub2 script sub script $ rpm -e test-sub test-sub2 sub2 script sub script -- Kacper From draenog at pld-linux.org Sat Mar 22 20:47:59 2014 From: draenog at pld-linux.org (Kacper Kornet) Date: Sat, 22 Mar 2014 20:47:59 +0100 Subject: [packages/poldek] - BR: libsemanage-devel libsemanage-static - added at-char-completion.patch - rel 6 In-Reply-To: <8cd07b070798719c37aeece598a1ee99c0674596_refs_heads_master@pld-linux.org> References: <0d17669a262f853fce0df37e144e3ce1dfb4de65_refs_heads_master@pld-linux.org> <8cd07b070798719c37aeece598a1ee99c0674596_refs_heads_master@pld-linux.org> Message-ID: <20140322194759.GA19654@camk.edu.pl> On Sat, Feb 15, 2014 at 09:35:42AM +0100, witekfl wrote: > commit 8cd07b070798719c37aeece598a1ee99c0674596 > Author: Witold Filipczyk > Date: Sat Feb 15 09:33:06 2014 +0100 > - BR: libsemanage-devel libsemanage-static > - added at-char-completion.patch > - rel 6 Where does BR: llibsemanage-devel come from? I'm asking as poldek and poldek-libs don't depend on libsemanage.so. What's more on my builder poldek builds fine without libsemanage-devel installed. -- Kacper From gglater62 at gmail.com Sat Mar 22 21:04:08 2014 From: gglater62 at gmail.com (Witold Filipczyk) Date: Sat, 22 Mar 2014 21:04:08 +0100 Subject: [packages/poldek] - BR: libsemanage-devel libsemanage-static - added at-char-completion.patch - rel 6 In-Reply-To: <20140322194759.GA19654@camk.edu.pl> References: <0d17669a262f853fce0df37e144e3ce1dfb4de65_refs_heads_master@pld-linux.org> <8cd07b070798719c37aeece598a1ee99c0674596_refs_heads_master@pld-linux.org> <20140322194759.GA19654@camk.edu.pl> Message-ID: W dniu 22.03.2014 o 20:47 Kacper Kornet pisze: > On Sat, Feb 15, 2014 at 09:35:42AM +0100, witekfl wrote: >> commit 8cd07b070798719c37aeece598a1ee99c0674596 >> Author: Witold Filipczyk >> Date: Sat Feb 15 09:33:06 2014 +0100 > >> - BR: libsemanage-devel libsemanage-static >> - added at-char-completion.patch >> - rel 6 > > Where does BR: llibsemanage-devel come from? I'm asking as > poldek and poldek-libs don't depend on libsemanage.so. What's more > on my builder poldek builds fine without libsemanage-devel installed. In the attachment buildlog of i686 (th, th-ready, th-test). I didn't think too much, but there was -lsemanage -- U?ywam klienta poczty Opera Mail: http://www.opera.com/mail/ -------------- next part -------------- A non-text attachment was scrubbed... Name: poldek.log.bz2 Type: application/bzip2 Size: 12671 bytes Desc: not available URL: From krzysztof.krakowiak at gmail.com Mon Mar 24 00:17:34 2014 From: krzysztof.krakowiak at gmail.com (Krzysztof Krakowiak) Date: Mon, 24 Mar 2014 00:17:34 +0100 Subject: [packages/poldek] - BR: libsemanage-devel libsemanage-static - added at-char-completion.patch - rel 6 Message-ID: On Sat, Feb 15, 2014 at 09:35:42AM +0100, witekfl wrote: > commit 8cd07b070798719c37aeece598a1ee99c0674596 > Author: Witold Filipczyk > Date: Sat Feb 15 09:33:06 2014 +0100 >> - BR: libsemanage-devel libsemanage-static >> - added at-char-completion.patch >> - rel 6 >Where does BR: llibsemanage-devel come from? I'm asking as >poldek and poldek-libs don't depend on libsemanage.so. What's more >on my builder poldek builds fine without libsemanage-devel installed. I don't know if its needed, but '*.so' files are dynamically linked libraries in ELF format, while 'BR' is about static library ('*.a' files). Usually when doing static linking, library is embedded in executable(not a separate file on hdd). From lukaszgl at post.pl Mon Mar 24 21:07:22 2014 From: lukaszgl at post.pl (Lukasz Glebicki) Date: Mon, 24 Mar 2014 21:07:22 +0100 Subject: /usr/bin/time for builder script ? Message-ID: <1434569.6k3v42S35G@inhell> Hi, After latest updates I had builder script (from repo and system) not working; open3: exec of time -p schedtool -B -e rpmbuild --target x86_64-linux -ba -v --define _specdir /home/users/builder/rpm/packages/amarok --define _sourcedir /home/users/builder/rpm/packages/amarok /home/users/builder/tmp/builder.amarok.Ijuhgc/amarok.spec failed at /home/users/builder/rpm/rpm-build-tools/teeboth line 35. or /usr/bin/builder: time: not found Where this dependency should be added? Regards, -- ?ukasz G??bicki mail/rot13:yhxnfmty at cbfg.cy PLD/Linux Team gg:246267 Linux Registered User #318551 blekot:{irc,skype} From glen at pld-linux.org Tue Mar 25 08:12:50 2014 From: glen at pld-linux.org (=?UTF-8?B?RWxhbiBSdXVzYW3DpGU=?=) Date: Tue, 25 Mar 2014 09:12:50 +0200 Subject: /usr/bin/time for builder script ? In-Reply-To: <1434569.6k3v42S35G@inhell> References: <1434569.6k3v42S35G@inhell> Message-ID: <53312C72.5020308@pld-linux.org> On 24/03/14 22:07, Lukasz Glebicki wrote: > Hi, > > After latest updates I had builder script (from repo and system) not working; > > open3: exec of time -p schedtool -B -e rpmbuild --target x86_64-linux -ba -v > --define _specdir /home/users/builder/rpm/packages/amarok --define _sourcedir > /home/users/builder/rpm/packages/amarok > /home/users/builder/tmp/builder.amarok.Ijuhgc/amarok.spec failed at > /home/users/builder/rpm/rpm-build-tools/teeboth line 35. > > or > > /usr/bin/builder: time: not found > > Where this dependency should be added? > rpm -qf /usr/bin/builder and it's already added: http://git.pld-linux.org/?p=packages/rpm-build-tools.git;a=commitdiff;h=1e555cc84178572f69fad547fa0e03eaead3cf19 -- glen From lukaszgl at post.pl Tue Mar 25 20:25:52 2014 From: lukaszgl at post.pl (Lukasz Glebicki) Date: Tue, 25 Mar 2014 20:25:52 +0100 Subject: /usr/bin/time for builder script ? In-Reply-To: <53312C72.5020308@pld-linux.org> References: <1434569.6k3v42S35G@inhell> <53312C72.5020308@pld-linux.org> Message-ID: <5153286.1ZAFPmvFZJ@inhell> On Tuesday 25 of March 2014 09:12:50 Elan Ruusam?e wrote: > rpm -qf /usr/bin/builder rpm-build-tools-4.6-2.noarch Release : 2 Build Date: wto, 4 mar 2014, 19:54:10 Install Date: pon, 24 mar 2014, 20:19:44 Build Host: ymir-builder Regarsd, -- ?ukasz G??bicki mail/rot13:yhxnfmty at cbfg.cy PLD/Linux Team gg:246267 Linux Registered User #318551 blekot:{irc,skype} From draenog at pld-linux.org Wed Mar 26 09:32:08 2014 From: draenog at pld-linux.org (Kacper Kornet) Date: Wed, 26 Mar 2014 09:32:08 +0100 Subject: [packages/poldek] - BR: libsemanage-devel libsemanage-static - added at-char-completion.patch - rel 6 In-Reply-To: References: <0d17669a262f853fce0df37e144e3ce1dfb4de65_refs_heads_master@pld-linux.org> <8cd07b070798719c37aeece598a1ee99c0674596_refs_heads_master@pld-linux.org> <20140322194759.GA19654@camk.edu.pl> Message-ID: <20140326083207.GB4124@camk.edu.pl> On Sat, Mar 22, 2014 at 09:04:08PM +0100, Witold Filipczyk wrote: > W dniu 22.03.2014 o 20:47 Kacper Kornet pisze: > >On Sat, Feb 15, 2014 at 09:35:42AM +0100, witekfl wrote: > >>commit 8cd07b070798719c37aeece598a1ee99c0674596 > >>Author: Witold Filipczyk > >>Date: Sat Feb 15 09:33:06 2014 +0100 > >> - BR: libsemanage-devel libsemanage-static > >> - added at-char-completion.patch > >> - rel 6 > >Where does BR: llibsemanage-devel come from? I'm asking as > >poldek and poldek-libs don't depend on libsemanage.so. What's more > >on my builder poldek builds fine without libsemanage-devel installed. > In the attachment buildlog of i686 (th, th-ready, th-test). > I didn't think too much, but there was -lsemanage Thanks, the buildlog was very useful. It appears that instead of BR in poldek.spec it should be R: in rpm-devel (and rpm-static for static libraries). What's more the build failed only on 32bit architectures as on x86_64 poldek used the old gettext macro which didn't know about /usr/lib64 and libtool didn't provide the linker with the dependant libraries. -- Kacper From glen at pld-linux.org Thu Mar 27 22:08:49 2014 From: glen at pld-linux.org (=?windows-1252?Q?Elan_Ruusam=E4e?=) Date: Thu, 27 Mar 2014 23:08:49 +0200 Subject: start_udev on udev removal Message-ID: <53349361.7070405@pld-linux.org> posting here, as maybe someday somebody wants to fix this. seems start_udev trigger is invoked when *udev* package is removed, and that call seems even to succeed! # rpm -q dev udev dev-3.4-10.x86_64 udev-208-11.x86_64 # rpm -e udev * Starting udev...................................................[ BUSY ] mknod: ?/dev/cpu/microcode?: File exists mknod: ?/dev/fuse?: File exists mknod: ?/dev/net/tun?: File exists mknod: ?/dev/ppp?: File exists mknod: ?/dev/mapper/control?: File exists [ DONE ] * Setup extra nodes...............................................[ DONE ] * Retrigger subsystems events.....................................[ DONE ] * Retrigger devices events........................................[ DONE ] * udevadm settle..................................................[ DONE ] # rpm -q udev dev package udev is not installed dev-3.4-10.x86_64 -- glen From draenog at pld-linux.org Thu Mar 27 23:03:04 2014 From: draenog at pld-linux.org (Kacper Kornet) Date: Thu, 27 Mar 2014 23:03:04 +0100 Subject: start_udev on udev removal In-Reply-To: <53349361.7070405@pld-linux.org> References: <53349361.7070405@pld-linux.org> Message-ID: <20140327220304.GA16913@camk.edu.pl> On Thu, Mar 27, 2014 at 11:08:49PM +0200, Elan Ruusam?e wrote: > posting here, as maybe someday somebody wants to fix this. > seems start_udev trigger is invoked when *udev* package is removed, > and that call seems even to succeed! > # rpm -q dev udev > dev-3.4-10.x86_64 > udev-208-11.x86_64 And how did you end with both dev and udev installed, as udev obsoletes dev? > # rpm -e udev > * Starting udev...................................................[ BUSY ] > mknod: ?/dev/cpu/microcode?: File exists > mknod: ?/dev/fuse?: File exists > mknod: ?/dev/net/tun?: File exists > mknod: ?/dev/ppp?: File exists > mknod: ?/dev/mapper/control?: File exists > [ DONE ] > * Setup extra nodes...............................................[ DONE ] > * Retrigger subsystems events.....................................[ DONE ] > * Retrigger devices events........................................[ DONE ] > * udevadm settle..................................................[ DONE ] > # rpm -q udev dev > package udev is not installed > dev-3.4-10.x86_64 It seems that the triggerpostun are run based on Provides. As udev provides dev so the trigger is run. It succeeds as start_udev belongs to udev-core so is still present in the system. The solution would be to limit the trigger to dev < 3.5. It is under assumption that dev proper will never reach that version. -- Kacper From glen at pld-linux.org Thu Mar 27 23:34:16 2014 From: glen at pld-linux.org (=?UTF-8?B?RWxhbiBSdXVzYW3DpGU=?=) Date: Fri, 28 Mar 2014 00:34:16 +0200 Subject: start_udev on udev removal In-Reply-To: <20140327220304.GA16913@camk.edu.pl> References: <53349361.7070405@pld-linux.org> <20140327220304.GA16913@camk.edu.pl> Message-ID: <5334A768.2000606@pld-linux.org> On 28/03/14 00:03, Kacper Kornet wrote: >> # rpm -q dev udev >> >dev-3.4-10.x86_64 >> >udev-208-11.x86_64 > And how did you end with both dev and udev installed, as udev obsoletes > dev? > installed "dev" after "udev*" was already installed, intention was to use static /dev -- glen From draenog at pld-linux.org Fri Mar 28 17:58:36 2014 From: draenog at pld-linux.org (Kacper Kornet) Date: Fri, 28 Mar 2014 17:58:36 +0100 Subject: dracut: ide vs ata modules Message-ID: <20140328165836.GA19301@camk.edu.pl> In PLD kernels there are both ide and ata drivers for the same controllers, for example ata_piix and piix. Dependant on which one is loaded first the disks are visible as /dev/sd* or /dev/hd*. The kernel help says "Users of ATA hardware are encouraged to migrate to the newer ATA subsystem ("Serial ATA (prod) and Parallel ATA (experimental) drivers") which is more actively maintained." Unfortunately as the dracut in the default configuration tries to load all matching modules in the order presented by modprobe, the ide modules are used in many cases. I can see three solutions: 1) Don't build the ide modules at all. It is the way Fedora went (they have CONFIG_IDE not set except on powerpc). However arekm claims that there are some ide modules not ported to ata. 2) Change the order of modules presented by modprobe for a given alias. That can be accomplished by a trivial patch to kernel Makefile. 3) Blacklist ide modules in dracut configuration I think the second option is the simplest one, so I would like to implement it in PLD, unless someone has preference. -- Kacper From arekm at maven.pl Fri Mar 28 18:50:09 2014 From: arekm at maven.pl (Arkadiusz =?utf-8?q?Mi=C5=9Bkiewicz?=) Date: Fri, 28 Mar 2014 18:50:09 +0100 Subject: dracut: ide vs ata modules In-Reply-To: <20140328165836.GA19301@camk.edu.pl> References: <20140328165836.GA19301@camk.edu.pl> Message-ID: <201403281850.09872.arekm@maven.pl> On Friday 28 of March 2014, Kacper Kornet wrote: > In PLD kernels there are both ide and ata drivers for the same controllers, > for example ata_piix and piix. Dependant on which one is loaded first > the disks are visible as /dev/sd* or /dev/hd*. > > The kernel help says "Users of ATA hardware are encouraged to migrate to > the newer ATA subsystem ("Serial ATA (prod) and Parallel ATA > (experimental) drivers") which is more actively maintained." > Unfortunately as the dracut in the default configuration tries to load > all matching modules in the order presented by modprobe, the ide modules > are used in many cases. > > I can see three solutions: > > 1) Don't build the ide modules at all. It is the way Fedora went (they > have CONFIG_IDE not set except on powerpc). However arekm claims that > there are some ide modules not ported to ata. > > 2) Change the order of modules presented by modprobe for a given alias. > That can be accomplished by a trivial patch to kernel Makefile. Fails to work with vanilla kernels where using vanilla kernel was supported configuration in pld. > I think the second option is the simplest one, so I would like to > implement it in PLD, unless someone has preference. 1) but maybe on per module basis? (unless my claim is not true then disable all) -- Arkadiusz Mi?kiewicz, arekm / maven.pl From hawk at pld-linux.org Fri Mar 28 21:22:41 2014 From: hawk at pld-linux.org (Marcin Krol) Date: Fri, 28 Mar 2014 21:22:41 +0100 Subject: dracut: ide vs ata modules In-Reply-To: <20140328165836.GA19301@camk.edu.pl> References: <20140328165836.GA19301@camk.edu.pl> Message-ID: <5335DA11.8090906@pld-linux.org> > I think the second option is the simplest one, so I would like to > implement it in PLD, unless someone has preference. Personally I'm using dracut with host-only option enabled. It uses system wide module blacklist and includes only modules for supported hardware in initramfs, not all of them. I've not encountered (yet) machine on which host-only config would prefer IDE drivers over SATA ones but perhaps it depends on hardware used. If host-only config is not enough then best solution IMO is to patch dracut to prefer SATA drivers over IDE. It can be done by checking device class if more than one module is available. M. From mike at osdn.org.ua Fri Mar 28 22:26:59 2014 From: mike at osdn.org.ua (Michael Shigorin) Date: Fri, 28 Mar 2014 23:26:59 +0200 Subject: dracut: ide vs ata modules In-Reply-To: <20140328165836.GA19301@camk.edu.pl> References: <20140328165836.GA19301@camk.edu.pl> Message-ID: <20140328212659.GR20822@osdn.org.ua> On Fri, Mar 28, 2014 at 05:58:36PM +0100, Kacper Kornet wrote: > I can see three solutions: > > 1) Don't build the ide modules at all. It is the way Fedora > went (they have CONFIG_IDE not set except on powerpc). However > arekm claims that there are some ide modules not ported to ata. > > 2) Change the order of modules presented by modprobe for a > given alias. That can be accomplished by a trivial patch to > kernel Makefile. > > 3) Blacklist ide modules in dracut configuration 4) Provide a separate subpackage, e.g. kernel-modules-ide-* in ALT Linux have been created exactly for that reason: so that those who actually need these can just install. -- ---- WBR, Michael Shigorin / http://altlinux.org ------ http://opennet.ru / http://anna-news.info From glen at pld-linux.org Sun Mar 30 12:41:13 2014 From: glen at pld-linux.org (=?UTF-8?B?RWxhbiBSdXVzYW3DpGU=?=) Date: Sun, 30 Mar 2014 13:41:13 +0300 Subject: [packages/vlc] vlc %post crashes, but allow package at least to install In-Reply-To: <6828f9c6544433caed4fe79e1a4b70d34203fc6d_refs_heads_master@pld-linux.org> References: <6828f9c6544433caed4fe79e1a4b70d34203fc6d_refs_heads_master@pld-linux.org> Message-ID: <5337F4C9.3020105@pld-linux.org> On 30.03.2014 13:40, glen wrote: > commit 6828f9c6544433caed4fe79e1a4b70d34203fc6d > Author: Elan Ruusam?e > Date: Sun Mar 30 13:40:09 2014 +0300 > > vlc %post crashes, but allow package at least to install ... > > %post > /sbin/ldconfig > -%{_libdir}/vlc/vlc-cache-gen -f %{_libdir}/vlc/plugins > +%{_libdir}/vlc/vlc-cache-gen -f %{_libdir}/vlc/plugins || : didn't this get fixed at some point? 22:vlc ########################################### [100%] Segmentation fault (core dumped) error: %post(vlc-2.1.2-2.x86_64) scriptlet failed, exit status 139 -- glen From baggins at pld-linux.org Sun Mar 30 21:41:08 2014 From: baggins at pld-linux.org (Jan =?utf-8?Q?R=C4=99korajski?=) Date: Sun, 30 Mar 2014 21:41:08 +0200 Subject: dracut: ide vs ata modules In-Reply-To: <20140328165836.GA19301@camk.edu.pl> References: <20140328165836.GA19301@camk.edu.pl> Message-ID: <20140330194108.GA1659@sith.mimuw.edu.pl> On Fri, 28 Mar 2014, Kacper Kornet wrote: > > In PLD kernels there are both ide and ata drivers for the same controllers, > for example ata_piix and piix. Dependant on which one is loaded first > the disks are visible as /dev/sd* or /dev/hd*. > > The kernel help says "Users of ATA hardware are encouraged to migrate to > the newer ATA subsystem ("Serial ATA (prod) and Parallel ATA > (experimental) drivers") which is more actively maintained." > Unfortunately as the dracut in the default configuration tries to load > all matching modules in the order presented by modprobe, the ide modules > are used in many cases. > > I can see three solutions: > > 1) Don't build the ide modules at all. It is the way Fedora went (they > have CONFIG_IDE not set except on powerpc). However arekm claims that > there are some ide modules not ported to ata. > > 2) Change the order of modules presented by modprobe for a given alias. That can be > accomplished by a trivial patch to kernel Makefile. > > 3) Blacklist ide modules in dracut configuration > > I think the second option is the simplest one, so I would like to > implement it in PLD, unless someone has preference. Personally, I'd just disable CONFIG_IDE, but if you say there is hardware not supported by the ata drivers I'm going to move IDE drivers to a subpackage. -- Jan R?korajski | PLD/Linux SysAdm | http://www.pld-linux.org/ bagginsmimuw.edu.pl bagginspld-linux.org