Re: running poudriere with 8 builders
On Saturday, 4 March 2017 17:19:12 CET, Michael Gmelinwrote: I now have made tar-backups of /usr/local/poudiere und /usr/local (without) poudriere; I will umount the 2nd disk (da1) and restore /usr/local, which will end up in the 1st disk; Make sure to initialize poudriere properly on zfs (it will create various filesystems for jails/ports etc). Simply untarring won't cut it. I moved /usr/local to the 1st disk, created a ZFS pool.on the other and there the jail and ports tree with the poudriere command... Now it's really starting fast, within a minute, and is building around 100 ports in 10 minutes, all CPUs run busy. thanks matthias -- Sent from my Ubuntu phone http://www.unixarea.de/ ___ freebsd-ports@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-ports To unsubscribe, send any mail to "freebsd-ports-unsubscr...@freebsd.org"
Re: running poudriere with 8 builders
Set maxvnodes to 50 and jump 100k until you're happy. I have Xeon e3 that uses 800,000 and thats reasonable; but i keep it at load 8 for a two days. Keep an eye on your swap, I set mine to swap out idle stuff early. Thanks for sharing your experience -- *Disclaimer:* *As implied by email protocols, the information in this message is not confidential. Any intermediary or recipient may inspect, modify (add), copy, forward, reply to, delete, or filter email for any purpose unless said parties are otherwise obligated. Nothing in this message may be legally binding without cryptographic evidence of its integrity and/or confidentiality.* ___ freebsd-ports@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-ports To unsubscribe, send any mail to "freebsd-ports-unsubscr...@freebsd.org"
Re: running poudriere with 8 builders
> On 4 Mar 2017, at 16:32, Matthias Apitzwrote: > >> El día Saturday, March 04, 2017 a las 12:57:40PM +0100, Kurt Jaeger escribió: >> >> Hi! >> Did you use ZFS as file system ? >>> >>> When I did the disk setup, I was also thinking in ZFS, but this would have >>> reduced the netto disk space from 2* 280 GB to the half. >> >> It does not have to be that way. Use only one disk for the system >> and the second disk for a poudriere ZFS pool ? >> >>> I have / for >>> all.the system and /usr/local for poudriere on the 2nd disk. >> >> Use the same setup, but with ZFS. > > I now have made tar-backups of /usr/local/poudiere und /usr/local > (without) poudriere; I will umount the 2nd disk (da1) and restore > /usr/local, which will end up in the 1st disk; > Make sure to initialize poudriere properly on zfs (it will create various filesystems for jails/ports etc). Simply untarring won't cut it. > about the ZFS I'm unsure; the poudriere.conf says: > > ... > #ZPOOL=tank > ... > # root of the poudriere zfs filesystem, by default /poudriere > # ZROOTFS=/poudriere > > Does this mean I should set: > > ZPOOL=poudriere > ZROOTFS=/poudriere > > and do: > > # zpool create poudriere /dev/da1 > > and restore the poudriere.tgz into this having > > /poudriere/poudriere/data/ > /poudriere/poudriere/jails/ > /poudriere/poudriere/ports/ > > or only > > /poudriere/data/ > /poudriere/jails/ > /poudriere/ports/ > > Thanks for clarification. > >matthias > > -- > Matthias Apitz, ✉ g...@unixarea.de, ⌂ http://www.unixarea.de/ ☎ > +49-176-38902045 > ___ > freebsd-ports@freebsd.org mailing list > https://lists.freebsd.org/mailman/listinfo/freebsd-ports > To unsubscribe, send any mail to "freebsd-ports-unsubscr...@freebsd.org" ___ freebsd-ports@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-ports To unsubscribe, send any mail to "freebsd-ports-unsubscr...@freebsd.org"
Re: running poudriere with 8 builders
Hi! > > Use the same setup, but with ZFS. > > I now have made tar-backups of /usr/local/poudiere und /usr/local > (without) poudriere; I will umount the 2nd disk (da1) and restore > /usr/local, which will end up in the 1st disk; > > about the ZFS I'm unsure; the poudriere.conf says: > > ... > #ZPOOL=tank > ... > # root of the poudriere zfs filesystem, by default /poudriere > # ZROOTFS=/poudriere > > Does this mean I should set: > > ZPOOL=poudriere > ZROOTFS=/poudriere I have ZPOOL=pou ZROOTFS=/pou (I'm too lazy to type /poudriere all the time 8-) > and do: > > # zpool create poudriere /dev/da1 > > and restore the poudriere.tgz into this having > > /poudriere/poudriere/data/ > /poudriere/poudriere/jails/ > /poudriere/poudriere/ports/ > > or only > > /poudriere/data/ > /poudriere/jails/ > /poudriere/ports/ The last one is correct. -- p...@opsec.eu+49 171 3101372 3 years to go ! ___ freebsd-ports@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-ports To unsubscribe, send any mail to "freebsd-ports-unsubscr...@freebsd.org"
Re: running poudriere with 8 builders
El día Saturday, March 04, 2017 a las 12:57:40PM +0100, Kurt Jaeger escribió: > Hi! > > > > Did you use ZFS as file system ? > > > > When I did the disk setup, I was also thinking in ZFS, but this would have > > reduced the netto disk space from 2* 280 GB to the half. > > It does not have to be that way. Use only one disk for the system > and the second disk for a poudriere ZFS pool ? > > > I have / for > > all.the system and /usr/local for poudriere on the 2nd disk. > > Use the same setup, but with ZFS. I now have made tar-backups of /usr/local/poudiere und /usr/local (without) poudriere; I will umount the 2nd disk (da1) and restore /usr/local, which will end up in the 1st disk; about the ZFS I'm unsure; the poudriere.conf says: ... #ZPOOL=tank ... # root of the poudriere zfs filesystem, by default /poudriere # ZROOTFS=/poudriere Does this mean I should set: ZPOOL=poudriere ZROOTFS=/poudriere and do: # zpool create poudriere /dev/da1 and restore the poudriere.tgz into this having /poudriere/poudriere/data/ /poudriere/poudriere/jails/ /poudriere/poudriere/ports/ or only /poudriere/data/ /poudriere/jails/ /poudriere/ports/ Thanks for clarification. matthias -- Matthias Apitz, ✉ g...@unixarea.de, ⌂ http://www.unixarea.de/ ☎ +49-176-38902045 ___ freebsd-ports@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-ports To unsubscribe, send any mail to "freebsd-ports-unsubscr...@freebsd.org"
Re: running poudriere with 8 builders
On Sat, Mar 04, 2017 at 12:57:40PM +0100 I heard the voice of Kurt Jaeger, and lo! it spake thus: > > /pou is on an Samsung SSD 840 EVO 250GB: > > smartctl says: > > 9 Power_On_Hours 0x0032 091 091 Always - 41885 > 12 Power_Cycle_Count 0x0032 099 099 Always - 8 > 177 Wear_Leveling_Count 0x0013 083 083 Always - 199 > > The Wear_Leveling_Count shows that it's probably beyond the spec > already 8-} Actually, I'm pretty sure it means the opposite 8-} It's my understanding that the "Raw" value on them is pretty much roughly the actual count of levelled drive writes, and the cooked "Value" (/worst, though that doesn't make a difference on that attribute) winds up basically meaning "percent of rated life left"[0]. So you've got ~200 drive writes, and have used about 17% of its life. On a 250GB 840 Evo I have, in somewhat easier use than being poundriered senseless: ID# ATTRIBUTE_NAME FLAGSVALUE WORST THRESH FAIL RAW_VALUE 177 Wear_Leveling_Count PO--C- 095 095 000-57 It's in a mirror with a similarly sized MX100, which (with its different attributes; can't be consistent, y'know) says 202 Percent_Lifetime_Used P---CK 095 095 000-5 which agrees; I've only got 19/20ths of their lifespan to go before I gotta cough up umpty-dozens of dollars to replace them. [0] For most SMART attributes, the cooked 'value' actually starts at 100 and counts down, with the Threshold level being where it starts claiming to be deeply unhappy. -- Matthew Fuller (MF4839) | fulle...@over-yonder.net Systems/Network Administrator | http://www.over-yonder.net/~fullermd/ On the Internet, nobody can hear you scream. ___ freebsd-ports@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-ports To unsubscribe, send any mail to "freebsd-ports-unsubscr...@freebsd.org"
Re: running poudriere with 8 builders
Kurt Jaeger wrote on 2017/03/04 12:57: /pou is on an Samsung SSD 840 EVO 250GB: smartctl says: 9 Power_On_Hours 0x0032 091 091 Always - 41885 12 Power_Cycle_Count 0x0032 099 099 Always - 8 177 Wear_Leveling_Count 0x0013 083 083 Always - 199 The Wear_Leveling_Count shows that it's probably beyond the spec already 8-} I don't think it's worn out. Wear Leveling starts at 100 and count down until there are 000. # smartctl -A /dev/ada2 | egrep 'ID|Wear_Level|Power' ID# ATTRIBUTE_NAME FLAG VALUE WORST THRESH TYPE UPDATED WHEN_FAILED RAW_VALUE 9 Power_On_Hours 0x0032 094 094 000Old_age Always - 28767 12 Power_Cycle_Count 0x0032 099 099 000Old_age Always - 6 177 Wear_Leveling_Count 0x0013 064 064 000Pre-fail Always - 1299 Miroslav Lachman ___ freebsd-ports@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-ports To unsubscribe, send any mail to "freebsd-ports-unsubscr...@freebsd.org"
Re: running poudriere with 8 builders
On 4 March 2017 at 13:20, Tommi Pernilawrote: > By default MAKE_JOBS is disabled to allow only one process per cpu. To > allow it anyway, ALLOW_MAKE_JOBS=yes in your /usr/local/etc/poudriere.conf: > > echo "ALLOW_MAKE_JOBS=yes" >> /usr/local/etc/poudriere.conf Consider allowing this only for selected ports, e.g. via setting ALLOW_MAKE_JOBS_PACKAGES="pkg gcc* py* " My machine tends to become close to unresponsive for (UI heavy tasks like browsing) when both firefox and chromium builds run in parallel with ALLOW_MAKE_JOBS. Riggs ___ freebsd-ports@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-ports To unsubscribe, send any mail to "freebsd-ports-unsubscr...@freebsd.org"
Re: running poudriere with 8 builders
On Mar 4, 2017 13:57, "Kurt Jaeger"wrote: Hi! > > Did you use ZFS as file system ? > > When I did the disk setup, I was also thinking in ZFS, but this would have > reduced the netto disk space from 2* 280 GB to the half. It does not have to be that way. Use only one disk for the system and the second disk for a poudriere ZFS pool ? > I have / for > all.the system and /usr/local for poudriere on the 2nd disk. Use the same setup, but with ZFS. I think, but have not checked it, that poudriere uses zfs clones, which is ***way*** more efficient than copying the whole tree for each poudriere builder jail. > > SSD instead of spinning disks ? > No. Seagate SAS disks. 3.5" or 2.5" ? Putting the poudriere storage on a SSD might be very helpful. This is my builder box (8 threads, Intel(R) Xeon(R) CPU E3-1230 V2 @ 3.30GHz, 32 GB RAM): zpool list NAMESIZE ALLOC FREE EXPANDSZ FRAGCAP DEDUP HEALTH ALTROOT zroot 428G 216G 212G -47%50% 1.00x ONLINE - pou 232G 49.9G 182G -39%21% 1.00x ONLINE - /pou is on an Samsung SSD 840 EVO 250GB: smartctl says: 9 Power_On_Hours 0x0032 091 091 Always - 41885 12 Power_Cycle_Count 0x0032 099 099 Always - 8 177 Wear_Leveling_Count 0x0013 083 083 Always - 199 The Wear_Leveling_Count shows that it's probably beyond the spec already 8-} -- p...@opsec.eu+49 171 3101372 3 years to go ! ___ freebsd-ports@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-ports To unsubscribe, send any mail to "freebsd-ports-unsubscr...@freebsd.org" Have you checked poudriere configurations MAKE_JOBS parameter? Here is a snippet from poudriere wiki: By default MAKE_JOBS is disabled to allow only one process per cpu. To allow it anyway, ALLOW_MAKE_JOBS=yes in your /usr/local/etc/poudriere.conf: echo "ALLOW_MAKE_JOBS=yes" >> /usr/local/etc/poudriere.conf Br,Tommi ___ freebsd-ports@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-ports To unsubscribe, send any mail to "freebsd-ports-unsubscr...@freebsd.org"
Re: running poudriere with 8 builders
Hi! > > Did you use ZFS as file system ? > > When I did the disk setup, I was also thinking in ZFS, but this would have > reduced the netto disk space from 2* 280 GB to the half. It does not have to be that way. Use only one disk for the system and the second disk for a poudriere ZFS pool ? > I have / for > all.the system and /usr/local for poudriere on the 2nd disk. Use the same setup, but with ZFS. I think, but have not checked it, that poudriere uses zfs clones, which is ***way*** more efficient than copying the whole tree for each poudriere builder jail. > > SSD instead of spinning disks ? > No. Seagate SAS disks. 3.5" or 2.5" ? Putting the poudriere storage on a SSD might be very helpful. This is my builder box (8 threads, Intel(R) Xeon(R) CPU E3-1230 V2 @ 3.30GHz, 32 GB RAM): zpool list NAMESIZE ALLOC FREE EXPANDSZ FRAGCAP DEDUP HEALTH ALTROOT zroot 428G 216G 212G -47%50% 1.00x ONLINE - pou 232G 49.9G 182G -39%21% 1.00x ONLINE - /pou is on an Samsung SSD 840 EVO 250GB: smartctl says: 9 Power_On_Hours 0x0032 091 091 Always - 41885 12 Power_Cycle_Count 0x0032 099 099 Always - 8 177 Wear_Leveling_Count 0x0013 083 083 Always - 199 The Wear_Leveling_Count shows that it's probably beyond the spec already 8-} -- p...@opsec.eu+49 171 3101372 3 years to go ! ___ freebsd-ports@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-ports To unsubscribe, send any mail to "freebsd-ports-unsubscr...@freebsd.org"
Re: running poudriere with 8 builders
On Saturday, 4 March 2017 12:33:39 CET, Kurt Jaegerwrote: Did you use ZFS as file system ? When I did the disk setup, I was also thinking in ZFS, but this would have reduced the netto disk space from 2* 280 GB to the half. I have / for all.the system and /usr/local for poudriere on the 2nd disk. SSD instead of spinning disks ? No. Seagate SAS disks. -- Sent from my Ubuntu phone http://www.unixarea.de/ ___ freebsd-ports@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-ports To unsubscribe, send any mail to "freebsd-ports-unsubscr...@freebsd.org"
Re: running poudriere with 8 builders
Hi! > During this time (00:09:20 - 00:33:47) the CPUs are idling and top > shows 8x 'cpdup' (I do not see any man page for this, what proc is > this?). > > The overall throughput now is around 70 ports per hour. And top shows > little CPU usage and a lot of pkg-static processes, perhaps for > unpacking dependency packages. > > Any hints about how to improve the setup? Did you use ZFS as file system ? SSD instead of spinning disks ? -- p...@opsec.eu+49 171 3101372 3 years to go ! ___ freebsd-ports@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-ports To unsubscribe, send any mail to "freebsd-ports-unsubscr...@freebsd.org"
running poudriere with 8 builders
Hello, I acquired a server which my company sent to dust (called it "decommissioning"), a Dell PowerEdge r210 with 8 CPU 3.3 GHz and 16 GByte RAM. I have had to insert 2 SAS disks, each of 300 GByte. The beast compiled CURRENT's world in an hour, and the kernel in 6 minutes, wow! Now I setup poudriere and a ports tree in it, all from CURRENT too. I started with 8 builders and poudriere needs halv an hour to start/clone the builders: root@jet:/usr/home/guru # poudriere bulk -f /usr/local/etc/poudriere-list -J 8 -j freebsd-r314251 -p ports-20170303 [00:00:00] >> Creating the reference jail... done [00:03:46] >> Mounting system devices for freebsd-r314251-ports-20170303 [00:03:46] >> Mounting ports/packages/distfiles [00:03:46] >> Using packages from previously failed build [00:03:46] >> Mounting packages from: /usr/local/poudriere/data/packages/freebsd-r314251-ports-20170303 [00:03:46] >> Copying /var/db/ports from: /usr/local/etc/poudriere.d/freebsd-r314251-options [00:03:47] >> Appending to make.conf: /usr/local/etc/poudriere.d/freebsd-r314251-make.conf /etc/resolv.conf -> /usr/local/poudriere/data/.m/freebsd-r314251-ports-20170303/ref/etc/resolv.conf [00:03:47] >> Starting jail freebsd-r314251-ports-20170303 [00:03:48] >> Logs: /usr/local/poudriere/data/logs/bulk/freebsd-r314251-ports-20170303/2017-03-04_09h50m18s [00:03:48] >> Loading MOVED [00:04:06] >> Calculating ports order and dependencies [00:09:17] >> pkg package missing, skipping sanity [00:09:17] >> Skipping incremental rebuild and repository sanity checks [00:09:17] >> Cleaning the build queue [00:09:20] >> Recording filesystem state for prepkg... done [00:09:20] >> Building 1760 packages using 8 builders [00:09:20] >> Starting/Cloning builders [00:33:47] >> Hit CTRL+t at any time to see build progress and stats [00:33:47] >> [01][00:00:00] Starting build of ports-mgmt/pkg [00:35:01] >> [01][00:01:14] Finished build of ports-mgmt/pkg: Success ... During this time (00:09:20 - 00:33:47) the CPUs are idling and top shows 8x 'cpdup' (I do not see any man page for this, what proc is this?). The overall throughput now is around 70 ports per hour. And top shows little CPU usage and a lot of pkg-static processes, perhaps for unpacking dependency packages. Any hints about how to improve the setup? Thanks matthias -- Matthias Apitz, ✉ g...@unixarea.de, ⌂ http://www.unixarea.de/ ☎ +49-176-38902045 ___ freebsd-ports@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-ports To unsubscribe, send any mail to "freebsd-ports-unsubscr...@freebsd.org"