On Tue, Sep 28, 2021 at 09:02:31PM +0200, Dick Visser wrote: > Begin: Running /scripts/local-premount ... done. > Begin: Will now check root file system ... fsck from util-linux 2.36.1 > [/sbin/fsck.ext4 (1) -- /dev/nvme0n1p1] fsck.ext4 -a -C0 /dev/nvme0n1p1 > /dev/nvme0n1p1: clean, 35442/516096 files, 242371/2064379 blocks > done. > [ 9.042449] EXT4-fs (nvme0n1p1): mounted filesystem with ordered > data mode. Opts: (null) > done. > Begin: Running /scripts/local-bottom ... [ 28.809447] nvme0n1: p1 p14 p15 > GROWROOT: CHANGED: partition=1 start=262144 old:[ 28.819359] > nvme0n1: p1 p14 p15 > size=16515039 end=16777183 new: size=41680863 end=41943007 > [ 40.228970] EXT4-fs (nvme0n1p1): mounted filesystem with ordered > data mode. Opts: (null) > done. > Begin: Running /scripts/init-bottom ... done. > [ 40.258874] Not activating Mandatory Access Control as > /sbin/tomoyo-init does not exist. > [ 40.359770] systemd[1]: Inserted module 'autofs4' That looks like unusually slow IO for some reason. Do you see reduced IO performance once the instance has finished booting, or is it only during the early boot period? Is there any correlation between availability zone and slow launches? Are they all clustered in one AZ? Assuming your EBS volumes are the gp2 type, it may be worth checking their BurstBalance metric. It seems unlikely that you'd be getting throttled on only ~10% of your launches, though, if that was it. See https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ebs-volume-types.html#EBSVolumeTypes_gp2 and https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ebs-io-characteristics.html#ebs-io-metrics for some details on the IO performance properties of EBS. noah
Attachment:
signature.asc
Description: PGP signature