fix(ci): Delete mainnet and testnet disks separately (#6582)

* Remove trailing whitespace in delete-gcp-resources.yml

* Document which delete-gcp-resources.yml steps already work for testnet

* Delete mainnet and testnet images separately

* Give shell variables unique names

Co-authored-by: Deirdre Connolly <durumcrustulum@gmail.com>

---------

Co-authored-by: Deirdre Connolly <durumcrustulum@gmail.com>
This commit is contained in:
teor 2023-04-28 12:26:29 +10:00 committed by GitHub
parent d375c3e9e5
commit 1f639ffff0
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 78 additions and 21 deletions

View File

@ -46,7 +46,7 @@ jobs:
- name: Set up Cloud SDK
uses: google-github-actions/setup-gcloud@v1.1.0
# Deletes all instances older than $DELETE_INSTANCE_DAYS days.
# Deletes all mainnet and testnet instances older than $DELETE_INSTANCE_DAYS days.
#
# We only delete instances that end in 7 or more hex characters,
# to avoid deleting managed instance groups and manually created instances.
@ -80,7 +80,7 @@ jobs:
gcloud compute instance-templates delete "${TEMPLATE}" || continue
done
# Deletes all the disks older than $DELETE_AGE_DAYS days.
# Deletes all mainnet and testnet disks older than $DELETE_AGE_DAYS days.
#
# Disks that are attached to an instance template can't be deleted, so it is safe to try to delete all disks here.
#
@ -102,7 +102,7 @@ jobs:
gcloud compute disks delete --verbosity=info ${DISK_AND_LOCATION} || continue
IFS=$'\n'
done
IFS=$'\n'
# Disks created by managed instance groups, and other jobs that start with "zebrad-"
ZEBRAD_DISKS=$(gcloud compute disks list --sort-by=creationTimestamp --filter="name~^zebrad- AND creationTimestamp < $DELETE_BEFORE_DATE" --format='value(NAME,LOCATION,LOCATION_SCOPE)' | \
@ -115,66 +115,123 @@ jobs:
IFS=$'\n'
done
# Deletes cache images older than $DELETE_AGE_DAYS days.
# Deletes mainnet and testnet cache images older than $DELETE_AGE_DAYS days.
#
# Keeps all images younger than $DELETE_AGE_DAYS.
# Also keeps $KEEP_LATEST_IMAGE_COUNT older images of each type:
# Also keeps $KEEP_LATEST_IMAGE_COUNT older images of each type, for each network:
# - zebrad checkpoint cache
# - zebrad tip cache
# - lightwalletd + zebrad tip cache
#
# TODO:
# - keep the latest $KEEP_LATEST_IMAGE_COUNT, if there are at least that many recent images, delete all the outdated images
# - when we add testnet to the workflows, keep the latest $KEEP_LATEST_IMAGE_COUNT testnet images,
# and the latest $KEEP_LATEST_IMAGE_COUNT mainnet images.
# - refactor out repeated shell script code
- name: Delete old cache disks
run: |
DELETE_BEFORE_DATE=$(date --date="$DELETE_AGE_DAYS days ago" '+%Y%m%d')
ZEBRAD_CHECKPOINT_IMAGES=$(gcloud compute images list --sort-by=~creationTimestamp --filter="name~^zebrad-cache-.*net-checkpoint AND creationTimestamp < $DELETE_BEFORE_DATE" --format='value(NAME)')
# As of April 2023, these disk names look like:
# zebrad-cache-6039-merge-62c8ecc-v25-mainnet-checkpoint-053559
#
# Mainnet zebrad checkpoint
ZEBRAD_MAINNET_CHECKPOINT_IMAGES=$(gcloud compute images list --sort-by=~creationTimestamp --filter="name~^zebrad-cache-.*-mainnet-checkpoint AND creationTimestamp < $DELETE_BEFORE_DATE" --format='value(NAME)')
KEPT_IMAGES=0
for IMAGE in $ZEBRAD_CHECKPOINT_IMAGES
for IMAGE in $ZEBRAD_MAINNET_CHECKPOINT_IMAGES
do
if [[ "$KEPT_IMAGES" -lt "$KEEP_LATEST_IMAGE_COUNT" ]];
then
then
KEPT_IMAGES=$((KEPT_IMAGES+1))
echo "Keeping image $KEPT_IMAGES named $IMAGE"
continue
fi
gcloud compute images delete "${IMAGE}" || continue
done
ZEBRAD_TIP_IMAGES=$(gcloud compute images list --sort-by=~creationTimestamp --filter="name~^zebrad-cache-.*net-tip AND creationTimestamp < $DELETE_BEFORE_DATE" --format='value(NAME)')
# Testnet zebrad checkpoint
ZEBRAD_TESTNET_CHECKPOINT_IMAGES=$(gcloud compute images list --sort-by=~creationTimestamp --filter="name~^zebrad-cache-.*-testnet-checkpoint AND creationTimestamp < $DELETE_BEFORE_DATE" --format='value(NAME)')
KEPT_IMAGES=0
for IMAGE in $ZEBRAD_TIP_IMAGES
for IMAGE in $ZEBRAD_TESTNET_CHECKPOINT_IMAGES
do
if [[ "$KEPT_IMAGES" -lt "$KEEP_LATEST_IMAGE_COUNT" ]];
then
then
KEPT_IMAGES=$((KEPT_IMAGES+1))
echo "Keeping image $KEPT_IMAGES named $IMAGE"
continue
fi
gcloud compute images delete "${IMAGE}" || continue
done
LWD_TIP_IMAGES=$(gcloud compute images list --sort-by=~creationTimestamp --filter="name~^lwd-cache-.*net-tip AND creationTimestamp < $DELETE_BEFORE_DATE" --format='value(NAME)')
# As of April 2023, these disk names look like:
# zebrad-cache-6556-merge-a2ca4de-v25-mainnet-tip(-u)?-140654
#
# Mainnet zebrad tip
ZEBRAD_MAINNET_TIP_IMAGES=$(gcloud compute images list --sort-by=~creationTimestamp --filter="name~^zebrad-cache-.*-mainnet-tip AND creationTimestamp < $DELETE_BEFORE_DATE" --format='value(NAME)')
KEPT_IMAGES=0
for IMAGE in $LWD_TIP_IMAGES
for IMAGE in $ZEBRAD_MAINNET_TIP_IMAGES
do
if [[ "$KEPT_IMAGES" -lt "$KEEP_LATEST_IMAGE_COUNT" ]];
then
then
KEPT_IMAGES=$((KEPT_IMAGES+1))
echo "Keeping image $KEPT_IMAGES named $IMAGE"
continue
fi
gcloud compute images delete "${IMAGE}" || continue
done
# Testnet zebrad tip
ZEBRAD_TESTNET_TIP_IMAGES=$(gcloud compute images list --sort-by=~creationTimestamp --filter="name~^zebrad-cache-.*-testnet-tip AND creationTimestamp < $DELETE_BEFORE_DATE" --format='value(NAME)')
KEPT_IMAGES=0
for IMAGE in $ZEBRAD_TESTNET_TIP_IMAGES
do
if [[ "$KEPT_IMAGES" -lt "$KEEP_LATEST_IMAGE_COUNT" ]];
then
KEPT_IMAGES=$((KEPT_IMAGES+1))
echo "Keeping image $KEPT_IMAGES named $IMAGE"
continue
fi
gcloud compute images delete "${IMAGE}" || continue
done
# As of April 2023, these disk names look like:
# lwd-cache-main-fb3fec0-v25-mainnet-tip(-u)?-061314
#
# Mainnet lightwalletd tip
LWD_MAINNET_TIP_IMAGES=$(gcloud compute images list --sort-by=~creationTimestamp --filter="name~^lwd-cache-.*-mainnet-tip AND creationTimestamp < $DELETE_BEFORE_DATE" --format='value(NAME)')
KEPT_IMAGES=0
for IMAGE in $LWD_MAINNET_TIP_IMAGES
do
if [[ "$KEPT_IMAGES" -lt "$KEEP_LATEST_IMAGE_COUNT" ]];
then
KEPT_IMAGES=$((KEPT_IMAGES+1))
echo "Keeping image $KEPT_IMAGES named $IMAGE"
continue
fi
gcloud compute images delete "${IMAGE}" || continue
done
# Testnet lightwalletd tip
LWD_TESTNET_TIP_IMAGES=$(gcloud compute images list --sort-by=~creationTimestamp --filter="name~^lwd-cache-.*-testnet-tip AND creationTimestamp < $DELETE_BEFORE_DATE" --format='value(NAME)')
KEPT_IMAGES=0
for IMAGE in $LWD_TESTNET_TIP_IMAGES
do
if [[ "$KEPT_IMAGES" -lt "$KEEP_LATEST_IMAGE_COUNT" ]];
then
KEPT_IMAGES=$((KEPT_IMAGES+1))
echo "Keeping image $KEPT_IMAGES named $IMAGE"
continue
fi
gcloud compute images delete "${IMAGE}" || continue
done
# We're using a generic approach here, which allows multiple registries to be included,
# even those not related to GCP. Enough reason to create a separate job.
#
# The same artifacts are used for both mainnet and testnet.
clean-registries:
name: Delete unused artifacts in registry
runs-on: ubuntu-latest