Increase [x86,amd64,arm]-generic image size |
|||||||
Issue descriptionhttps://uberchromegw.corp.google.com/i/chromeos/builders/x86-generic-chromium-pfq/builds/8547 Too many graphics drivers plus component build bust the image. Also increase arm and amd64-generic sizes. Right now from 1.2GB to maybe 2GB straight? (Generic only.) 22859776 usr/lib/dri/i915_dri.so 22859776 usr/lib/dri/kms_swrast_dri.so 22859776 usr/lib/dri/nouveau_dri.so 22859776 usr/lib/dri/r300_dri.so 22859776 usr/lib/dri/r600_dri.so 22859776 usr/lib/dri/swrast_dri.so 24948736 opt/google/chrome/lib/libwebcore_shared.so 27422720 opt/google/chrome/lib/libcontent.so 30806016 opt/google/chrome/pepper/libpepflashplayer.so 60432384 opt/google/chrome/chrome ERROR : Target image has run out of space: ERROR : Filesystem 1M-blocks Used Available Use% Mounted on ERROR : /dev/loop0 1205 1205 0 100% /mnt/host/source/src/build/images/x86-generic/R52-8244.0.0-rc1/rootfs
,
Apr 26 2016
Pri -> 1 since this is blocking the PFQ.
,
Apr 26 2016
I would propose to add a disk_layout.json for the generic boards and override the ROOT fs_size. deymo@ - do you have any context why our ROOT-A partition has a size of 2048 MiB but only a fs_size of 1224 MiB? Is there any problem to bump that to, say, 1536 MiB?
,
Apr 26 2016
the fs_size is the size of the formatted filesystem, while the other one is the partition size. You want the filesystem size to be as small as possible (basically, just the size of your data) because that's what every update will ship. If you make the fs_size 1.9G but only ship 1.1G of data, then every update will have to re-write ~800M of zeros (to make dm-verity happy). The payload size of an extra ~800M of zeros is negligible (like maybe 2 or 3 Kb more of payload) but the I/O required to write those is not, and I/O is kind of the bottleneck during the update. I/O also makes your device slow because it disrupts other queued writes (chrome likes to do fsync on some "important files"). If production images on most chromebooks are gonna have 1.5G of data, then sure, bump it to 1.5G (although my next question would be: do you really need to add 300M to our builds??); but if this is for some particular internal-only setup then bump the size of those boards only. The .json files are stackable. The partition size is 2G so we have room to ship more stuff in the future; but meanwhile we don't need that much overhead in the update.
,
Apr 26 2016
We are discussing *generic images here, not production. I don't see a reason why we can't bump these only all the way to 2GB and forget about them. No reason to ever waste engineering time on this process in the future.
,
Apr 26 2016
,
Apr 26 2016
While we're at it, let's increase all generic board's rootfs size
,
Apr 26 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromiumos/overlays/board-overlays/+/f77eef6006a3bd49221b9ca1c39ef242825082e8 commit f77eef6006a3bd49221b9ca1c39ef242825082e8 Author: Haixia Shi <hshi@chromium.org> Date: Tue Apr 26 20:33:08 2016 disk_layout.json: Increase ROOT-A fs_size to 2000 MiB for generic boards. These generic boards are for internal testing only and not for production. They also require larger rootfs because of the extra drivers required. Therefore we shall just increase the file system size to the maximum allowed by the partition size and not waste more time thereafter. BUG= chromium:606862 TEST=build_images --board=[x86,arm,amd64]-generic and verify ROOT-A size Change-Id: I1ebddf929b46a5b27c1cfdd87a227f27fb267a04 Reviewed-on: https://chromium-review.googlesource.com/340832 Reviewed-by: Ilja Friedel <ihf@chromium.org> Trybot-Ready: Ilja Friedel <ihf@chromium.org> Tested-by: Ilja Friedel <ihf@chromium.org> Commit-Queue: Haixia Shi <hshi@chromium.org> [add] https://crrev.com/f77eef6006a3bd49221b9ca1c39ef242825082e8/overlay-amd64-generic/scripts/disk_layout.json [add] https://crrev.com/f77eef6006a3bd49221b9ca1c39ef242825082e8/overlay-arm-generic/scripts/disk_layout.json [add] https://crrev.com/f77eef6006a3bd49221b9ca1c39ef242825082e8/overlay-x86-generic/scripts/disk_layout.json
,
Apr 26 2016
,
Apr 27 2016
|
|||||||
►
Sign in to add a comment |
|||||||
Comment 1 by h...@chromium.org
, Apr 26 2016