ceph/qa/workunits/cephadm/test_cephadm.sh

   1 #!/bin/bash -ex
   2
   3 SCRIPT_NAME=$(basename ${BASH_SOURCE[0]})
   4 SCRIPT_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
   5
   6 # cleanup during exit
   7 [ -z "$CLEANUP" ] && CLEANUP=true
   8
   9 FSID='00000000-0000-0000-0000-0000deadbeef'
  10
  11 # images that are used
  12 IMAGE_MASTER=${IMAGE_MASTER:-'quay.ceph.io/ceph-ci/ceph:master'}
  13 IMAGE_PACIFIC=${IMAGE_PACIFIC:-'quay.ceph.io/ceph-ci/ceph:pacific'}
  14 #IMAGE_OCTOPUS=${IMAGE_OCTOPUS:-'quay.ceph.io/ceph-ci/ceph:octopus'}
  15 IMAGE_DEFAULT=${IMAGE_MASTER}
  16
  17 OSD_IMAGE_NAME="${SCRIPT_NAME%.*}_osd.img"
  18 OSD_IMAGE_SIZE='6G'
  19 OSD_TO_CREATE=2
  20 OSD_VG_NAME=${SCRIPT_NAME%.*}
  21 OSD_LV_NAME=${SCRIPT_NAME%.*}
  22
  23 CEPHADM_SRC_DIR=${SCRIPT_DIR}/../../../src/cephadm
  24 CEPHADM_SAMPLES_DIR=${CEPHADM_SRC_DIR}/samples
  25
  26 [ -z "$SUDO" ] && SUDO=sudo
  27
  28 if [ -z "$CEPHADM" ]; then
  29     CEPHADM=${CEPHADM_SRC_DIR}/cephadm
  30 fi
  31
  32 # at this point, we need $CEPHADM set
  33 if ! [ -x "$CEPHADM" ]; then
  34     echo "cephadm not found. Please set \$CEPHADM"
  35     exit 1
  36 fi
  37
  38 # add image to args
  39 CEPHADM_ARGS="$CEPHADM_ARGS --image $IMAGE_DEFAULT"
  40
  41 # combine into a single var
  42 CEPHADM_BIN="$CEPHADM"
  43 CEPHADM="$SUDO $CEPHADM_BIN $CEPHADM_ARGS"
  44
  45 # clean up previous run(s)?
  46 $CEPHADM rm-cluster --fsid $FSID --force
  47 $SUDO vgchange -an $OSD_VG_NAME || true
  48 loopdev=$($SUDO losetup -a | grep $(basename $OSD_IMAGE_NAME) | awk -F : '{print $1}')
  49 if ! [ "$loopdev" = "" ]; then
  50     $SUDO losetup -d $loopdev
  51 fi
  52
  53 # TMPDIR for test data
  54 [ -d "$TMPDIR" ] || TMPDIR=$(mktemp -d tmp.$SCRIPT_NAME.XXXXXX)
  55 [ -d "$TMPDIR_TEST_MULTIPLE_MOUNTS" ] || TMPDIR_TEST_MULTIPLE_MOUNTS=$(mktemp -d tmp.$SCRIPT_NAME.XXXXXX)
  56
  57 function cleanup()
  58 {
  59     if [ $CLEANUP = false ]; then
  60         # preserve the TMPDIR state
  61         echo "========================"
  62         echo "!!! CLEANUP=$CLEANUP !!!"
  63         echo
  64         echo "TMPDIR=$TMPDIR"
  65         echo "========================"
  66         return
  67     fi
  68
  69     dump_all_logs $FSID
  70     rm -rf $TMPDIR
  71 }
  72 trap cleanup EXIT
  73
  74 function expect_false()
  75 {
  76         set -x
  77         if eval "$@"; then return 1; else return 0; fi
  78 }
  79
  80 function is_available()
  81 {
  82     local name="$1"
  83     local condition="$2"
  84     local tries="$3"
  85
  86     local num=0
  87     while ! eval "$condition"; do
  88         num=$(($num + 1))
  89         if [ "$num" -ge $tries ]; then
  90             echo "$name is not available"
  91             false
  92         fi
  93         sleep 5
  94     done
  95
  96     echo "$name is available"
  97     true
  98 }
  99
 100 function dump_log()
 101 {
 102     local fsid="$1"
 103     local name="$2"
 104     local num_lines="$3"
 105
 106     if [ -z $num_lines ]; then
 107         num_lines=100
 108     fi
 109
 110     echo '-------------------------'
 111     echo 'dump daemon log:' $name
 112     echo '-------------------------'
 113
 114     $CEPHADM logs --fsid $fsid --name $name -- --no-pager -n $num_lines
 115 }
 116
 117 function dump_all_logs()
 118 {
 119     local fsid="$1"
 120     local names=$($CEPHADM ls | jq -r '.[] | select(.fsid == "'$fsid'").name')
 121
 122     echo 'dumping logs for daemons: ' $names
 123     for name in $names; do
 124         dump_log $fsid $name
 125     done
 126 }
 127
 128 function nfs_stop()
 129 {
 130     # stop the running nfs server
 131     local units="nfs-server nfs-kernel-server"
 132     for unit in $units; do
 133         if systemctl status $unit < /dev/null; then
 134             $SUDO systemctl stop $unit
 135         fi
 136     done
 137
 138     # ensure the NFS port is no longer in use
 139     expect_false "$SUDO ss -tlnp '( sport = :nfs )' | grep LISTEN"
 140 }
 141
 142 ## prepare + check host
 143 $SUDO $CEPHADM check-host
 144
 145 ## run a gather-facts (output to stdout)
 146 $SUDO $CEPHADM gather-facts
 147
 148 ## version + --image
 149 $SUDO CEPHADM_IMAGE=$IMAGE_PACIFIC $CEPHADM_BIN version
 150 $SUDO CEPHADM_IMAGE=$IMAGE_PACIFIC $CEPHADM_BIN version \
 151     | grep 'ceph version 16'
 152 #$SUDO CEPHADM_IMAGE=$IMAGE_OCTOPUS $CEPHADM_BIN version
 153 #$SUDO CEPHADM_IMAGE=$IMAGE_OCTOPUS $CEPHADM_BIN version \
 154 #    | grep 'ceph version 15'
 155 $SUDO $CEPHADM_BIN --image $IMAGE_MASTER version | grep 'ceph version'
 156
 157 # try force docker; this won't work if docker isn't installed
 158 systemctl status docker > /dev/null && ( $CEPHADM --docker version | grep 'ceph version' ) || echo "docker not installed"
 159
 160 ## test shell before bootstrap, when crash dir isn't (yet) present on this host
 161 $CEPHADM shell --fsid $FSID -- ceph -v | grep 'ceph version'
 162 $CEPHADM shell --fsid $FSID -e FOO=BAR -- printenv | grep FOO=BAR
 163
 164 # test stdin
 165 echo foo | $CEPHADM shell -- cat | grep -q foo
 166
 167 ## bootstrap
 168 ORIG_CONFIG=`mktemp -p $TMPDIR`
 169 CONFIG=`mktemp -p $TMPDIR`
 170 MONCONFIG=`mktemp -p $TMPDIR`
 171 KEYRING=`mktemp -p $TMPDIR`
 172 IP=127.0.0.1
 173 cat <<EOF > $ORIG_CONFIG
 174 [global]
 175         log to file = true
 176         osd crush chooseleaf type = 0
 177 EOF
 178 $CEPHADM bootstrap \
 179       --mon-id a \
 180       --mgr-id x \
 181       --mon-ip $IP \
 182       --fsid $FSID \
 183       --config $ORIG_CONFIG \
 184       --output-config $CONFIG \
 185       --output-keyring $KEYRING \
 186       --output-pub-ssh-key $TMPDIR/ceph.pub \
 187       --allow-overwrite \
 188       --skip-mon-network \
 189       --skip-monitoring-stack
 190 test -e $CONFIG
 191 test -e $KEYRING
 192 rm -f $ORIG_CONFIG
 193
 194 $SUDO test -e /var/log/ceph/$FSID/ceph-mon.a.log
 195 $SUDO test -e /var/log/ceph/$FSID/ceph-mgr.x.log
 196
 197 for u in ceph.target \
 198              ceph-$FSID.target \
 199              ceph-$FSID@mon.a \
 200              ceph-$FSID@mgr.x; do
 201     systemctl is-enabled $u
 202     systemctl is-active $u
 203 done
 204 systemctl | grep system-ceph | grep -q .slice  # naming is escaped and annoying
 205
 206 # check ceph -s works (via shell w/ passed config/keyring)
 207 $CEPHADM shell --fsid $FSID --config $CONFIG --keyring $KEYRING -- \
 208       ceph -s | grep $FSID
 209
 210 for t in mon mgr node-exporter prometheus grafana; do
 211     $CEPHADM shell --fsid $FSID --config $CONFIG --keyring $KEYRING -- \
 212              ceph orch apply $t --unmanaged
 213 done
 214
 215 ## ls
 216 $CEPHADM ls | jq '.[]' | jq 'select(.name == "mon.a").fsid' \
 217     | grep $FSID
 218 $CEPHADM ls | jq '.[]' | jq 'select(.name == "mgr.x").fsid' \
 219     | grep $FSID
 220
 221 # make sure the version is returned correctly
 222 $CEPHADM ls | jq '.[]' | jq 'select(.name == "mon.a").version' | grep -q \\.
 223
 224 ## deploy
 225 # add mon.b
 226 cp $CONFIG $MONCONFIG
 227 echo "public addrv = [v2:$IP:3301,v1:$IP:6790]" >> $MONCONFIG
 228 $CEPHADM deploy --name mon.b \
 229       --fsid $FSID \
 230       --keyring /var/lib/ceph/$FSID/mon.a/keyring \
 231       --config $MONCONFIG
 232 for u in ceph-$FSID@mon.b; do
 233     systemctl is-enabled $u
 234     systemctl is-active $u
 235 done
 236 cond="$CEPHADM shell --fsid $FSID --config $CONFIG --keyring $KEYRING -- \
 237             ceph mon stat | grep '2 mons'"
 238 is_available "mon.b" "$cond" 30
 239
 240 # add mgr.y
 241 $CEPHADM shell --fsid $FSID --config $CONFIG --keyring $KEYRING -- \
 242       ceph auth get-or-create mgr.y \
 243       mon 'allow profile mgr' \
 244       osd 'allow *' \
 245       mds 'allow *' > $TMPDIR/keyring.mgr.y
 246 $CEPHADM deploy --name mgr.y \
 247       --fsid $FSID \
 248       --keyring $TMPDIR/keyring.mgr.y \
 249       --config $CONFIG
 250 for u in ceph-$FSID@mgr.y; do
 251     systemctl is-enabled $u
 252     systemctl is-active $u
 253 done
 254
 255 for f in `seq 1 30`; do
 256     if $CEPHADM shell --fsid $FSID \
 257              --config $CONFIG --keyring $KEYRING -- \
 258           ceph -s -f json-pretty \
 259         | jq '.mgrmap.num_standbys' | grep -q 1 ; then break; fi
 260     sleep 1
 261 done
 262 $CEPHADM shell --fsid $FSID --config $CONFIG --keyring $KEYRING -- \
 263       ceph -s -f json-pretty \
 264     | jq '.mgrmap.num_standbys' | grep -q 1
 265
 266 # add osd.{1,2,..}
 267 dd if=/dev/zero of=$TMPDIR/$OSD_IMAGE_NAME bs=1 count=0 seek=$OSD_IMAGE_SIZE
 268 loop_dev=$($SUDO losetup -f)
 269 $SUDO vgremove -f $OSD_VG_NAME || true
 270 $SUDO losetup $loop_dev $TMPDIR/$OSD_IMAGE_NAME
 271 $SUDO pvcreate $loop_dev && $SUDO vgcreate $OSD_VG_NAME $loop_dev
 272
 273 # osd boostrap keyring
 274 $CEPHADM shell --fsid $FSID --config $CONFIG --keyring $KEYRING -- \
 275       ceph auth get client.bootstrap-osd > $TMPDIR/keyring.bootstrap.osd
 276
 277 # create lvs first so ceph-volume doesn't overlap with lv creation
 278 for id in `seq 0 $((--OSD_TO_CREATE))`; do
 279     $SUDO lvcreate -l $((100/$OSD_TO_CREATE))%VG -n $OSD_LV_NAME.$id $OSD_VG_NAME
 280 done
 281
 282 for id in `seq 0 $((--OSD_TO_CREATE))`; do
 283     device_name=/dev/$OSD_VG_NAME/$OSD_LV_NAME.$id
 284     CEPH_VOLUME="$CEPHADM ceph-volume \
 285                        --fsid $FSID \
 286                        --config $CONFIG \
 287                        --keyring $TMPDIR/keyring.bootstrap.osd --"
 288
 289     # prepare the osd
 290     $CEPH_VOLUME lvm prepare --bluestore --data $device_name --no-systemd
 291     $CEPH_VOLUME lvm batch --no-auto $device_name --yes --no-systemd
 292
 293     # osd id and osd fsid
 294     $CEPH_VOLUME lvm list --format json $device_name > $TMPDIR/osd.map
 295     osd_id=$($SUDO cat $TMPDIR/osd.map | jq -cr '.. | ."ceph.osd_id"? | select(.)')
 296     osd_fsid=$($SUDO cat $TMPDIR/osd.map | jq -cr '.. | ."ceph.osd_fsid"? | select(.)')
 297
 298     # deploy the osd
 299     $CEPHADM deploy --name osd.$osd_id \
 300           --fsid $FSID \
 301           --keyring $TMPDIR/keyring.bootstrap.osd \
 302           --config $CONFIG \
 303           --osd-fsid $osd_fsid
 304 done
 305
 306 # add node-exporter
 307 ${CEPHADM//--image $IMAGE_DEFAULT/} deploy \
 308     --name node-exporter.a --fsid $FSID
 309 cond="curl 'http://localhost:9100' | grep -q 'Node Exporter'"
 310 is_available "node-exporter" "$cond" 10
 311
 312 # add prometheus
 313 cat ${CEPHADM_SAMPLES_DIR}/prometheus.json | \
 314         ${CEPHADM//--image $IMAGE_DEFAULT/} deploy \
 315             --name prometheus.a --fsid $FSID --config-json -
 316 cond="curl 'localhost:9095/api/v1/query?query=up'"
 317 is_available "prometheus" "$cond" 10
 318
 319 # add grafana
 320 cat ${CEPHADM_SAMPLES_DIR}/grafana.json | \
 321         ${CEPHADM//--image $IMAGE_DEFAULT/} deploy \
 322             --name grafana.a --fsid $FSID --config-json -
 323 cond="curl --insecure 'https://localhost:3000' | grep -q 'grafana'"
 324 is_available "grafana" "$cond" 50
 325
 326 # add nfs-ganesha
 327 nfs_stop
 328 nfs_rados_pool=$(cat ${CEPHADM_SAMPLES_DIR}/nfs.json | jq -r '.["pool"]')
 329 $CEPHADM shell --fsid $FSID --config $CONFIG --keyring $KEYRING -- \
 330         ceph osd pool create $nfs_rados_pool 64
 331 $CEPHADM shell --fsid $FSID --config $CONFIG --keyring $KEYRING -- \
 332         rados --pool nfs-ganesha --namespace nfs-ns create conf-nfs.a
 333 $CEPHADM shell --fsid $FSID --config $CONFIG --keyring $KEYRING -- \
 334          ceph orch pause
 335 $CEPHADM deploy --name nfs.a \
 336       --fsid $FSID \
 337       --keyring $KEYRING \
 338       --config $CONFIG \
 339       --config-json ${CEPHADM_SAMPLES_DIR}/nfs.json
 340 cond="$SUDO ss -tlnp '( sport = :nfs )' | grep 'ganesha.nfsd'"
 341 is_available "nfs" "$cond" 10
 342 $CEPHADM shell --fsid $FSID --config $CONFIG --keyring $KEYRING -- \
 343          ceph orch resume
 344
 345 # add alertmanager via custom container
 346 alertmanager_image=$(cat ${CEPHADM_SAMPLES_DIR}/custom_container.json | jq -r '.image')
 347 tcp_ports=$(cat ${CEPHADM_SAMPLES_DIR}/custom_container.json | jq -r '.ports | map_values(.|tostring) | join(" ")')
 348 cat ${CEPHADM_SAMPLES_DIR}/custom_container.json | \
 349       ${CEPHADM//--image $IMAGE_DEFAULT/} \
 350       --image $alertmanager_image \
 351       deploy \
 352       --tcp-ports "$tcp_ports" \
 353       --name container.alertmanager.a \
 354       --fsid $FSID \
 355       --config-json -
 356 cond="$CEPHADM enter --fsid $FSID --name container.alertmanager.a -- test -f \
 357       /etc/alertmanager/alertmanager.yml"
 358 is_available "alertmanager.yml" "$cond" 10
 359 cond="curl 'http://localhost:9093' | grep -q 'Alertmanager'"
 360 is_available "alertmanager" "$cond" 10
 361
 362 ## run
 363 # WRITE ME
 364
 365 ## unit
 366 $CEPHADM unit --fsid $FSID --name mon.a -- is-enabled
 367 $CEPHADM unit --fsid $FSID --name mon.a -- is-active
 368 expect_false $CEPHADM unit --fsid $FSID --name mon.xyz -- is-active
 369 $CEPHADM unit --fsid $FSID --name mon.a -- disable
 370 expect_false $CEPHADM unit --fsid $FSID --name mon.a -- is-enabled
 371 $CEPHADM unit --fsid $FSID --name mon.a -- enable
 372 $CEPHADM unit --fsid $FSID --name mon.a -- is-enabled
 373
 374 ## shell
 375 $CEPHADM shell --fsid $FSID -- true
 376 $CEPHADM shell --fsid $FSID -- test -d /var/log/ceph
 377 expect_false $CEPHADM --timeout 10 shell --fsid $FSID -- sleep 60
 378 $CEPHADM --timeout 60 shell --fsid $FSID -- sleep 10
 379 $CEPHADM shell --fsid $FSID --mount $TMPDIR $TMPDIR_TEST_MULTIPLE_MOUNTS -- stat /mnt/$(basename $TMPDIR)
 380
 381 ## enter
 382 expect_false $CEPHADM enter
 383 $CEPHADM enter --fsid $FSID --name mon.a -- test -d /var/lib/ceph/mon/ceph-a
 384 $CEPHADM enter --fsid $FSID --name mgr.x -- test -d /var/lib/ceph/mgr/ceph-x
 385 $CEPHADM enter --fsid $FSID --name mon.a -- pidof ceph-mon
 386 expect_false $CEPHADM enter --fsid $FSID --name mgr.x -- pidof ceph-mon
 387 $CEPHADM enter --fsid $FSID --name mgr.x -- pidof ceph-mgr
 388 # this triggers a bug in older versions of podman, including 18.04's 1.6.2
 389 #expect_false $CEPHADM --timeout 5 enter --fsid $FSID --name mon.a -- sleep 30
 390 $CEPHADM --timeout 60 enter --fsid $FSID --name mon.a -- sleep 10
 391
 392 ## ceph-volume
 393 $CEPHADM ceph-volume --fsid $FSID -- inventory --format=json \
 394       | jq '.[]'
 395
 396 ## preserve test state
 397 [ $CLEANUP = false ] && exit 0
 398
 399 ## rm-daemon
 400 # mon and osd require --force
 401 expect_false $CEPHADM rm-daemon --fsid $FSID --name mon.a
 402 # mgr does not
 403 $CEPHADM rm-daemon --fsid $FSID --name mgr.x
 404
 405 expect_false $CEPHADM zap-osds --fsid $FSID
 406 $CEPHADM zap-osds --fsid $FSID --force
 407
 408 ## rm-cluster
 409 expect_false $CEPHADM rm-cluster --fsid $FSID --zap-osds
 410 $CEPHADM rm-cluster --fsid $FSID --force --zap-osds
 411
 412 echo PASS