Обсуждение: Reminder: spin up REL_11_STABLE on your buildfarm animals
Please check that your buildfarm members are building the v11 branch, and update their configurations if not. We currently have ~80 animals building v10, but only about 50 have checked in on v11 so far. regards, tom lane
On 07/07/2018 03:49 PM, Tom Lane wrote: > Please check that your buildfarm members are building the v11 branch, > and update their configurations if not. We currently have ~80 animals > building v10, but only about 50 have checked in on v11 so far. > > Yeah, here's the list: pgbfprod=> select sysname, operating_system, os_version, compiler, compiler_version from public.dashboard_mat where branch= 'REL_10_STABLE' and sysname not in (select sysname from public.dashboard_mat where branch = 'REL_11_STABLE'); sysname | operating_system | os_version | compiler | compiler_version ---------------+-------------------------+------------------------+---------------+-------------------------------------------- aholehole | CentOS | 5 | GCC | 4.1.2 brolga | Cygwin/XP-Pro | 1.7.7 | gcc | 4.3.4 calliphoridae | Debian | sid | gcc | 6.3 castoroides | Solaris | 10 | Sun Studio | 12 chub | RedHat Enterprise Linux | 7 | gcc | gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-4) conchuela | DragonFly BSD | 4.4.3-RELEASE | gcc | 5.2.1 culicidae | Debian | sid | gcc | 6.3 curculio | OpenBSD | 5.9 | gcc | 4.2.1 currawong | Windows XP-PRO | SP3 | MSVC++ | 2008 Express dory | Windows Server | 2016 | Visual Studio | 2015 flaviventris | Debian | Sid | gcc | snapshot francolin | Debian | sid | gcc | 6.3 frogmouth | Windows XP/Pro | SP3 | gcc | 4.5.0 gaur | HP-UX | 10.20 | gcc | 2.95.3 grison | Raspbian GNU/Linux | 7.8 | gcc | 4.6.3 guaibasaurus | Debian | 8.1 | gcc | 4.9.2 hamerkop | Windows | Server 2008 R2 (64bit) | Visual C++ | 2005 leech | CentOS Linux | 6.2 | icc | 14.0 loach | FreeBSD | 10.3-RELEASE | clang | 3.4.1 macaque | RHEL | 7.0 | gcc | 4.8.2 mule | Debian | Debian 9.0 | gcc | gcc 6.3.0 mylodon | Debian | sid | clang | 5.0 pademelon | HP-UX | 10.20 | HP C Compiler | A.10.32.22 piculet | Debian | sid | gcc | 6.3 protosciurus | Solaris | 10 | GCC | 3.4.3 rhinoceros | CentOS Linux | 7.1 | gcc | 4.8.3 20140911 (Red Hat 4.8.3-9) serinus | Debian | Sid | gcc | snapshot sidewinder | NetBSD | 7 | gcc | 4.8.4 skink | Debian Sid | sid | gcc | gcc-6.3 snapper | Debian | 7 Wheezy | gcc | 4.7 tick | CentOS Linux | 6.2 | clang/llvm | 3.4 Note that brolga, currawong and frogmouth will not build past Release 10 because the platform doesn't support huge pages, so they are expected to be on this list. That leaves 28. If you're running the modern way via run_branches.pl with branches_to_build set to ALL or HEAD_PLUS_LATESTn then this should just happen without owner intervention when a new branch is created. So I assume that all these are either running not using run_branches.pl. or have hardcoded lists of branches in their config files. Probably not best practice. cheers andrew -- Andrew Dunstan https://www.2ndQuadrant.com PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
On 2018-07-08 17:52, Andrew Dunstan wrote: > If you're running the modern way via run_branches.pl with > branches_to_build set to ALL or HEAD_PLUS_LATESTn then this should just > happen without owner intervention when a new branch is created. So I > assume that all these are either running not using run_branches.pl. or > have hardcoded lists of branches in their config files. Yes, I am using the run_build.pl on cron-schedule as I run most of the my animals on the same virtualized host and I need to control when different animal/branches builds so they don't overlap. If I would run run_branches.pl I have no control over how long each run will take and it's hard to schedule. But I've updated all my animals to include V11 now so they should all report in on the next run. > Probably not best practice. Probably not. But see above. /Mikael
On 07/09/2018 05:35 AM, Mikael Kjellström wrote: > > On 2018-07-08 17:52, Andrew Dunstan wrote: > >> If you're running the modern way via run_branches.pl with >> branches_to_build set to ALL or HEAD_PLUS_LATESTn then this should >> just happen without owner intervention when a new branch is created. >> So I assume that all these are either running not using >> run_branches.pl. or have hardcoded lists of branches in their config >> files. > > Yes, I am using the run_build.pl on cron-schedule as I run most of the > my animals on the same virtualized host and I need to control when > different animal/branches builds so they don't overlap. If I would > run run_branches.pl I have no control over how long each run will take > and it's hard to schedule. > > But I've updated all my animals to include V11 now so they should all > report in on the next run. > run_branches has a global lock that protects against concurrent runs of different animals. You just point all the animals on the machine at the same lock directory. I have several machines with this setup. See global_lock_dir in the sample config file. cheers andrew -- Andrew Dunstan https://www.2ndQuadrant.com PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
On 2018-07-09 13:31, Andrew Dunstan wrote: > run_branches has a global lock that protects against concurrent runs of > different animals. You just point all the animals on the machine at the > same lock directory. I have several machines with this setup. See > global_lock_dir in the sample config file. Yes, that would work if the animals run in the same OS-instance/virtual machine. All my animals is running different OS (OpenBSD, FreeBSD, NetBSD, DragonFlyBSD etc.) and in different virtual machines. /Mikael
On 07/09/2018 08:09 AM, Mikael Kjellström wrote: > On 2018-07-09 13:31, Andrew Dunstan wrote: > >> run_branches has a global lock that protects against concurrent runs >> of different animals. You just point all the animals on the machine >> at the same lock directory. I have several machines with this setup. >> See global_lock_dir in the sample config file. > > Yes, that would work if the animals run in the same > OS-instance/virtual machine. > > All my animals is running different OS (OpenBSD, FreeBSD, NetBSD, > DragonFlyBSD etc.) and in different virtual machines. > > Fair enough. I'll make a note that we need to notify buildfarm owners explicitly. cheers andrew -- Andrew Dunstan https://www.2ndQuadrant.com PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services