Another round of moving core cloudbender user-data feature into the base image

kubezero/kubezero/APKBUILD

@@ -0,0 +1,35 @@
+# Contributor: Stefan Reimer <stefan@zero-downtime.net>
+# Maintainer: Stefan Reimer <stefan@zero-downtime.net>
+pkgname=kubezero
+pkgver=1.24
+pkgrel=0
+pkgdesc="KubeZero release package"
+url="https://git.zero-downtime.net/ZeroDownTime/alpine-overlay/src/branch/master/kubezero/kubezero"
+arch="noarch"
+license="AGPL-3.0"
+depends="
+        cri-tools
+        cri-o=~$pkgver
+        kubelet=~$pkgver
+        kubectl=~$pkgver
+        ecr-credential-provider=~$pkgver
+        aws-iam-authenticator=~0.5.9
+        "
+options="!check"
+
+source="
+       shared-sys-fs.start
+       "
+
+build() {
+  return 0
+}
+
+package() {
+  # core library 
+  install -Dm755 "$srcdir"/shared-sys-fs.start "$pkgdir/etc/local.d/shared-sys-fs.start"
+}
+
+sha512sums="
+b0cadf577ea912630efabf8d104f2edaa79bd1697a1f9224ce8a75354dd204196c6d3c15c0318afa44be10be9696ce20ef0015198ee0b74050897d164f77ae60  shared-sys-fs.start
+"

kubezero/kubezero/shared-sys-fs.start

@@ -0,0 +1,3 @@
+#!/bin/sh
+mount --make-shared /sys/fs/cgroup
+mount --make-shared /sys

kubezero/zdt-base/APKBUILD

@@ -7,13 +7,15 @@ pkgdesc="ZeroDownTime Alpine additions and customizations"
 url="https://git.zero-downtime.net/ZeroDownTime/alpine-overlay/src/branch/master/kubezero/zdt-base"
 arch="noarch"
 license="AGPL-3.0"
-depends="logrotate syslog-ng neofetch monit"
+depends="logrotate syslog-ng neofetch monit file"
 options="!check"
 subpackages="$pkgname-openrc $pkgname-aws"
 install="$pkgname.post-install"
 
 source="
-        zdt-mount.init
+        lib-base.sh
+        cb-mount-var.init
+        cb-volumes.startstop
         zdt-sysctl.conf
         https://raw.githubusercontent.com/pixelb/ps_mem/v3.14/ps_mem.py
         syslog-ng.conf
@@ -33,15 +35,23 @@ build() {
 }
 
 package() {
+  # core library 
+  install -Dm755 "$srcdir/lib-base.sh" "$pkgdir/usr/lib/cloudbender/base.sh"
+
   # dhcp tuning for MTU
   install -Dm644 "$srcdir"/dhclient.conf "$pkgdir"/etc/dhcp/dhclient.conf
 
   # various sysctl tunings
   install -Dm644 "$srcdir"/zdt-sysctl.conf "$pkgdir"/etc/sysctl.d/60-zdt.conf
 
-  # init script to find and mount /var
-  mkdir -p "$pkgdir"/etc/init.d
-  cp zdt-mount.init "$pkgdir"/etc/init.d/zdt-mount
+  # init script to mount var as early as possible, cannot use any network !
+	install -Dm755 "$srcdir/cb-mount-var.init" "$pkgdir/etc/init.d/cb-mount-var"
+
+  # ensure "local" init script runs before user-data
+  mkdir -p "$pkgdir/etc/conf.d" 
+  echo 'rc_before="tiny-cloud-final"' > "$pkgdir/etc/conf.d/local"
+	install -Dm755 "$srcdir/cb-volumes.startstop" "$pkgdir/etc/local.d/cb-volumes.start"
+	( cd $pkgdir/etc/local.d; ln -s cb-volumes.start cb-volumes.stop; )
 
   # syslog-ng configs, json all into messages
   install -Dm644 "$srcdir"/syslog-ng.conf "$pkgdir"/lib/zdt/syslog-ng.conf
@@ -74,7 +84,9 @@ aws() {
 }
 
 sha512sums="
-16f4020e2e1f93b13b2ce140dea0c31066a55709cb3ae2ece54b9a6db57583e226bc43ac62be18f5a60274b87ae0de8c6bc613597988451853cdf085cae245eb  zdt-mount.init
+62e5bd982d3e957ca445891b00cc9fcdc3df22414cd332321a6046ae4ee4c98f9646d3680d83a6d643f01ded229bfea6f968e5734a58a5d233ac899c92ce85da  lib-base.sh
+0d78bb09b143576b1bc582a62868236e4febed306aa9d085570e91cf9cfbc77dd379342ade9f99203d822f830bbd55d42dcba52cb934952c7b749e252fab1eb3  cb-mount-var.init
+b4fbbf55c1a4d38c2877bade1d5e2ce5f1276a6704b0bb95b025e66a7c678710a60a8d4f37cb1f136af1435657cd4ffd03709e80fb61f8950ee39520c1a47f31  cb-volumes.startstop
 b9479835d8667fa99f8b1b140f969f0464a9bb3c60c7d19b57e306cfe82357d453932791e446caded71fddd379161ae8328367f1ee75ae3afc1b85e12294b621  zdt-sysctl.conf
 76e6a4f309f31bfa07de2d3b1faebe5670722752e18157b69d6e868cbe9e85eda393aed0728b0347a01a810eee442844c78259f86ff71e3136a013f4cbfaaea4  ps_mem.py
 9d087f2d4403a9c6d4d2f06fbb86519f2b8b134d8eb305facaef07c183815f917fb7bac916d39d504dbab7fdf3321a3f70954dde57e8986cc223371715bb1c54  syslog-ng.conf
@@ -84,7 +96,7 @@ b928ba547af080a07dc9063d44cb0f258d0e88e7c5a977e8f1cf1263c23608f0a138b8ffca0cdf58
 346b0170ada6cc1207ffb7b8ef138a1570a63c7df4d57618aa4b6b6c0d2df2197b0f5b23578ec83c641ee5e724865ac06985222e125809c990467426a0851b72  neofetch.conf
 532b8e2eb04942ab20bdc36b5dea1c60239fcbfcb85706123f3e05c18d65c938b85e9072d964ae5793177625a8db47b532db1f5bd5ed5ecbb70d5a331666ff54  zdt-ascii.txt
 c565516121b9e6f9d5f769511eb900546753e67cc4208d1b388fdce44cd28699261a5c3905f9a168d4b2d45ac65ac3a2a6a95335f1bbd76d2f444d5f50ec5c9e  dhclient.conf
-399356eaf09b41cde101aa9164eb492dc824e4bc75d8cd2197d1c2d6120349462dad2791609fb073285b3d3545067611f4608ff14b9d9586a46909269f496c02  cloudbender.stop
+cd7ddd7923d45370275fa26c0f2c6dea930c6788c8f55af4388eb42309125c15e5cbb34b186ab4aebbeac3470bed0ba2db9dd46ba8796242b59092f51c5cedf5  cloudbender.stop
 2d419d5c25a3829e99326b09876f459e48ab66f5756a8ad39b406c0f2829f5a323e8ff512afd8f32b7b07f24c88efa911bee495ce6c4d1925194cb54d3ba57bd  route53.py
 00eaff6c0a506580340b2547c3b1602a54238bac6090a15516839411478a4b4fdc138668b8ad23455445131f3a3e3fda175ed4bb0dd375402641c0e7b69c3218  get_iam_sshkeys.py
 "

kubezero/zdt-base/cb-mount-var.init

@@ -0,0 +1,20 @@
+#!/sbin/openrc-run
+# vim:set ts=8 noet ft=sh:
+
+description="Find suitable block device, prepare and mount it under /var"
+
+depend() {
+	need fsck root
+	use lvm modules
+	after clock lvm modules
+	before bootmisc
+}
+
+start() {
+	source /usr/lib/cloudbender/base.sh
+
+	ebegin "Looking for suitable /var"
+	setup_var
+	eend $?
+}
+

kubezero/zdt-base/cb-volumes.startstop

@@ -0,0 +1,18 @@
+#!/bin/sh
+# vim:set ts=8 noet ft=sh:
+#
+. /usr/lib/cloudbender/base.sh
+
+if [ "${0##*/}" = cb-volumes.start ]; then
+	get_meta_data
+	[ -z "$volumes" ] && return 0
+
+	mount_volumes "$volumes"
+
+elif [ "${0##*/}" = cb-volumes.stop ]; then
+
+	get_meta_data
+	[ -z "$volumes" ] && return 0
+
+	unmount_volumes "$volumes"
+fi

kubezero/zdt-base/cloudbender.stop

@@ -10,6 +10,6 @@ done
 [ $DEBUG -eq 1 ] && SHUTDOWNLOG="$(cat /tmp/shutdown.log)"
 
 [ -n "$RC_REBOOT" ] && ACTION="rebooting" || ACTION="terminated"
-[ -z "$DISABLE_SCALING_EVENTS" ] && cloudbender_sns_alarm.sh "Instance $ACTION" "" Info "$SHUTDOWNLOG"
+[ -z "$DISABLE_SCALING_EVENTS" ] && /var/lib/cloudbender/sns_alarm.sh "Instance $ACTION" "" Info "$SHUTDOWNLOG"
 
 sleep ${SHUTDOWN_PAUSE:-0}

kubezero/zdt-base/etc/init.d/zdt-mount

@@ -1,16 +0,0 @@
-#!/sbin/openrc-run
-# vim:set ts=8 noet ft=sh:
-
-description="ZDT stateful /var"
-
-depend() {
-	after mdev
-	before syslog-ng
-}
-
-start() {
-	ebegin "Looking for suitable /var"
-	echo "fake it"
-	eend $?
-}
-

kubezero/zdt-base/lib-base.sh

@@ -0,0 +1,136 @@
+#!/bin/sh
+
+# We built on top of tiny-cloud
+. /etc/conf.d/tiny-cloud
+
+# extract user-data args and cloud meta-data into /var/lib/cloud/meta-data
+get_meta_data() {
+  if [ ! -f /var/lib/cloud/meta-data ]; then
+  	[ -f /var/lib/cloud/user-data ] && bash /var/lib/cloud/user-data get_meta_data || echo "Error trying to extract cloud meta-data" >&2
+  fi
+
+  . /var/lib/cloud/meta-data
+}
+
+# archive orig /var, mount new var on top and restore orig var
+copy_and_mount() {
+  local dev=$1
+
+  tar cf /tmp/var.tar /var 2>/dev/null
+  mount -t xfs -o noatime "$dev" /var
+  tar xf /tmp/var.tar -C / && rm -f /tmp/var.tar
+}
+
+setup_var() {
+  for d in $(find /dev/sd?); do
+    # resolve to a valid block device
+    dev="$(realpath "$d")"
+    [ -b "$dev" ] || continue
+
+    # already mounted
+    mount | grep -q "$dev" && continue
+
+    case "$CLOUD" in
+      aws)
+        # on AWS look for sdx 
+        if [ "$d" = "/dev/sdx" ]; then
+          # check volume for existing filesystem
+          type=$(file -Lbs $d)
+          if [[ "$type" =~ "XFS filesystem" ]]; then
+            xfs_repair $d >/dev/null 2>&1
+          else
+            mkfs.xfs -qf $d >/dev/null
+          fi
+          copy_and_mount "$dev"
+          grep -q "$dev" /etc/fstab || echo "$dev /var xfs defaults,noatime,nofail 0 2" >> /etc/fstab
+        fi
+        ;;
+      *)
+        echo "Unsupported Cloud '$CLOUD'" >&2
+        exit 1
+        ;;
+    esac
+  done
+}
+
+attach_ebs() {
+  local volId="$1"
+  local device="$2"
+
+  local tries=30
+  while true; do
+    _json="$(aws ec2 describe-volumes --volume-ids $volId --region $REGION --output json)"
+    rc=$?; [ $rc -ne 0 ] && return $rc
+
+    vol_status=$(echo "$_json" | jq -r .Volumes[].State)
+    attachId=$(echo "$_json" | jq -r .Volumes[].Attachments[].InstanceId)
+
+    [ "$attachId" = "$INSTANCE_ID" ] && break
+
+    if [ "$vol_status" = "available" ]; then
+      aws ec2 attach-volume --volume-id "$volId" --instance-id "$INSTANCE_ID" --region "$REGION" --device "$device" > /dev/null
+      rc=$?; [ $rc -ne 0 ] && return $rc
+      break
+    fi
+
+    # if attached but not to us -> detach
+    if [ "$vol_status" = "in-use" ]; then
+      aws ec2 detach-volume --volume-id "$volId" --region "$REGION" --force
+      rc=$?; [ $rc -ne 0 ] && return $rc
+    fi
+
+    ((tries=tries-1))
+    [ $tries -eq 0 ] && return 1
+    sleep 5
+  done
+}
+
+_parse_volume() {
+  # Todo: proper checks once all is yaml
+  # For now just replace ':'
+  echo $1 | sed -e 's/:/ /g'
+}
+
+# mount optional remote volumes
+mount_volumes() {
+  local volumes="$1"
+
+  for vol in $volumes; do
+    # Todo: check volume type and call matching func
+    read volType volId volDevice volPath < <(_parse_volume $vol)
+
+    [ "$volType" != "ebs" ] && { echo "Unknown volume type $volType"; break; }
+    attach_ebs $volId $volDevice
+    rc=$?
+    [ $rc -ne 0 ] && { ewarn "error trying to attach $volId"; break; }
+
+    # wait for the block device to become available
+    while true; do
+      mdev -s
+      test -b $volDevice && break
+      sleep 1
+    done
+
+    # check volume for existing filesystem
+    type=$(file -Lbs $volDevice)
+    if [[ "$type" =~ "XFS filesystem" ]]; then
+      xfs_repair $volDevice >/dev/null 2>&1
+    else
+      mkfs.xfs -qf $volDevice >/dev/null
+    fi
+
+    # mount
+    mkdir -p $volPath
+    mount -t xfs -o noatime $volDevice $volPath
+  done
+}
+
+unmount_volumes() {
+  local volumes="$1"
+
+  for vol in $volumes; do
+    read volType volId volDevice volPath < <(_parse_volume $vol)
+
+    umount $volPath && aws ec2 detach-volume --volume-id "$volId" --instance-id $INSTANCE_ID --region $REGION > /dev/null
+  done
+}

kubezero/zdt-base/zdt-base.post-install

@@ -4,9 +4,14 @@
 sed -i -e 's/^[\s#]*TCPKeepAlive\s.*/TCPKeepAlive yes/' -e 's/^[\s#]*ClientAliveInterval\s.*/ClientAliveInterval 60/' /etc/ssh/sshd_config
 echo 'enabled SSH keep alives'
 
-# CgroupsV2
+# openRC
 sed -i -e 's/^[\s#]*rc_cgroup_mode=.*/rc_cgroup_mode="unified"/' /etc/rc.conf
-echo 'enabled cgroupv2'
+sed -i -e 's/^[\s#]*rc_logger=.*/rc_logger="YES"/' /etc/rc.conf
+echo 'enabled cgroupv2, openRC logging'
+
+# OpenRC parallel - causes too much chaos
+#sed -i -e 's/^[\s#]*rc_parallel=.*/rc_parallel="YES"/' /etc/rc.conf
+#echo 'enable parallel openRC'
 
 # Setup syslog-ng json logging
 cp /lib/zdt/syslog-ng.conf /etc/syslog-ng/syslog-ng.conf

kubezero/zdt-base/zdt-mount.init

@@ -1,16 +0,0 @@
-#!/sbin/openrc-run
-# vim:set ts=8 noet ft=sh:
-
-description="ZDT stateful /var"
-
-depend() {
-	after mdev
-	before syslog-ng
-}
-
-start() {
-	ebegin "Looking for suitable /var"
-	echo "fake it"
-	eend $?
-}
-