Repository navigation
Migration v26.1
This guide is for administrators moving a Docker Compose installation from ReportPortal 26.0.7 to 26.1. Read Breaking changes first. Docker Compose does not copy objects and does not rewrite paths in PostgreSQL.
- Pick your path
- Breaking changes
- What changed
- Option A: Filesystem storage
- Option B: S3-compatible storage
- Option C: Amazon S3
- Safety and rollback
- Known limits
Start from the 26.0.7 docker-compose.yml. Storage settings in that file are inside x-environment (API, Jobs, UAT) and again inside x-analyzer-environment. There is no x-storage block. Then pick the option that matches the storage you run today:
| Your 26.0.7 install | What you do | Where |
|---|---|---|
Filesystem (DATASTORE_TYPE: filesystem, volume storage at /data/storage) |
Take the 26.1 file. x-storage is already filesystem. No object copy |
Option A |
External MinIO, SeaweedFS, Ceph RGW, or another S3 API (DATASTORE_TYPE: minio, or s3 with your own endpoint), multi
|
Move the settings into x-storage as s3-compatible and copy the buckets into folders inside the default bucket |
Option B |
The same external store with RP_FEATURE_FLAGS: singleBucket
|
Move the settings into x-storage as s3-compatible. No copy |
Option B |
Amazon S3 (DATASTORE_TYPE: s3, RP_FEATURE_FLAGS: singleBucket) |
Move the settings into x-storage as aws-s3. No copy |
Option C |
Read this section before you change anything. Each item can make existing files unreachable or stop the release from starting.
- Storage settings moved to
x-storage. In 26.0.7,DATASTORE_TYPEand the otherDATASTORE_*keys live inx-environment, and the analyzer repeats them inx-analyzer-environment. The 26.1 file removes those keys from both blocks and adds one anchor,x-storage, which both blocks include. A value you leave in the old place is not applied. Copy your values intox-storagebefore you start the new release. See Where the settings moved. -
DATASTORE_TYPE: minioandDATASTORE_TYPE: s3are no longer valid. Uses3-compatible(MinIO, SeaweedFS, Ceph RGW, and other S3 APIs) oraws-s3.filesystemis unchanged. - The 26.0.7 S3 example spells the secret
DATASTORE_SECRETSKEY. The variable the services read isDATASTORE_SECRETKEY. A key left under the old spelling is ignored, and the API starts without your secret. - The
multiobject layout changed. In 26.0,multimeans one bucket per project (prj-<projectId>). In 26.1,multikeeps the same names but as folders inside the default bucket (rp-bucket/prj-<projectId>/…). 26.1 does not read the 26.0 buckets. Existing data must be copied into the new layout. Filesystem storage is not affected. See Where objects live. - The UAT service is removed.
service-apinow issues tokens and manages sessions. The token endpoint moves from/uat/sso/oauth/tokento/api/sso/oauth/token.docker compose up -d --remove-orphansremoves the olduatcontainer. - The
indexservice is removed. The same--remove-orphansflag removes that container. - The superadmin login changes. In 26.1 the built-in
superadminaccount is renamed toadmin@reportportal.internal. Sign in with the new name. Scripts that usesuperadminas the login must change too.
Compared releases:
| 26.0.7 | 26.1 | |
|---|---|---|
| Service images |
5.15.x (API_IMAGE, JOBS_IMAGE, UI_IMAGE, MIGRATIONS_IMAGE, ANALYZER_IMAGE) |
6.0.0 on the same variables |
| Storage settings | Inside x-environment and again in x-analyzer-environment
|
One x-storage anchor, included by both |
DATASTORE_TYPE |
minio, s3, filesystem
|
s3-compatible, aws-s3, filesystem
|
| API and Jobs filesystem path | Volume mount /data/storage
|
DATASTORE_PATH: /data/storage plus the same mount. The analyzer does not read DATASTORE_PATH
|
| Analyzer filesystem path | FILESYSTEM_DEFAULT_PATH: /data/storage/analyzer |
Unchanged. This is the path service-auto-analyzer 6.0.0 reads |
| Analyzer store type |
ANALYZER_BINARYSTORE_TYPE and a separate DATASTORE_TYPE
|
DATASTORE_TYPE from x-storage. ANALYZER_BINARYSTORE_TYPE is deprecated |
| Secret variable in the S3 example | DATASTORE_SECRETSKEY |
DATASTORE_SECRETKEY |
| Default bucket mode |
multi, unless RP_FEATURE_FLAGS: singleBucket
|
multi (folders in one bucket). singleBucket is unchanged |
| Analyzer object path | DATASTORE_DEFAULTBUCKETNAME: <bucket>/analyzer |
DATASTORE_BUCKETPREFIX: analyzer/prj- on the shared bucket |
| Superadmin login | superadmin |
admin@reportportal.internal |
| Token endpoint | /uat/sso/oauth/token |
/api/sso/oauth/token |
service-api, service-jobs, and service-auto-analyzer use Apache OpenDAL and accept only s3-compatible, aws-s3, and filesystem.
| 26.0.7 value | 26.1 value |
|---|---|
DATASTORE_TYPE: minio |
DATASTORE_TYPE: s3-compatible |
DATASTORE_TYPE: s3 |
DATASTORE_TYPE: aws-s3 |
DATASTORE_TYPE: filesystem |
unchanged. The analyzer reads this value |
ANALYZER_BINARYSTORE_TYPE |
deprecated. Remove it. 6.0.0 warns when it is set, and uses DATASTORE_TYPE
|
RP_FEATURE_FLAGS: singleBucket still selects one bucket with the 26.0 key layout. Leave it unset for multi.
In Docker Compose, DATASTORE_ENDPOINT keeps the scheme and the port, for example http://minio:9000.
The commented MinIO secret in 26.0.7 is miniopassword. The commented secret in 26.1 is rppassword. If you set STORAGE_SECRETKEY in .env, that value still wins. If you relied on the compose default, write the password you use today into DATASTORE_SECRETKEY. A new default does not open the old store.
26.0.7 keeps storage on the services that include x-environment. The analyzer does not use that anchor. It has its own DATASTORE_TYPE.
# 26.0.7 — inside x-environment (API, Jobs, UAT)
x-environment: &common-environment
# ... database and AMQP ...
DATASTORE_TYPE: filesystem # or s3
# 26.0.7 — inside x-analyzer-environment (analyzer only)
x-analyzer-environment: &common-analyzer-environment
ANALYZER_BINARYSTORE_TYPE: filesystem
DATASTORE_TYPE: filesystem # or minio, or s3
FILESYSTEM_DEFAULT_PATH: &analyzer-default-path /data/storage/analyzer26.1 defines DATASTORE_TYPE once. x-environment and x-analyzer-environment both pull the anchor in with <<: *storage-environment. service-auto-analyzer 6.0.0 reads DATASTORE_TYPE. ANALYZER_BINARYSTORE_TYPE is deprecated: if it is still set, the process warns and keeps DATASTORE_TYPE. Delete the old variable. The analyzer does not read DATASTORE_PATH. Its filesystem directory stays FILESYSTEM_DEFAULT_PATH.
# 26.1
x-storage: &storage-environment
DATASTORE_TYPE: filesystem
DATASTORE_PATH: /data/storage # API and Jobs only
x-environment: &common-environment
# ... database and AMQP ...
<<: *storage-environment
x-analyzer-environment: &common-analyzer-environment
<<: *storage-environment # DATASTORE_TYPE comes from here
FILESYSTEM_DEFAULT_PATH: &analyzer-default-path /data/storage/analyzerBefore you replace docker-compose.yml, copy every DATASTORE_* line and RP_FEATURE_FLAGS you changed in both blocks. After you take the 26.1 file:
- Put the API and Jobs values into
x-storageonly. - Leave
x-environmentas database, AMQP, and<<: *storage-environment. - Delete
ANALYZER_BINARYSTORE_TYPEand any secondDATASTORE_TYPEfromx-analyzer-environment. - Keep
FILESYSTEM_DEFAULT_PATHthere. That is the analyzer filesystem path.DATASTORE_PATHdoes not replace it. - For object storage, uncomment
DATASTORE_BUCKETPREFIX: analyzer/prj-inx-analyzer-environment. That replaces the 26.0.7 analyzer bucket name<bucket>/analyzer. The prefix applies only to the analyzer, because it is set after the merge.
DATASTORE_PATH is new, and only API and Jobs read it. It must be the path where the storage volume is mounted on those containers. The 26.1 default is /data/storage, which matches storage:/data/storage. If you replaced that mount with a bind such as ./data/storage:/data/storage, put the same bind back on api and jobs and keep DATASTORE_PATH equal to the container path. The analyzer volume stays mounted at FILESYSTEM_DEFAULT_PATH (/data/storage/analyzer).
In 26.1 the API opens one bucket, DATASTORE_DEFAULTBUCKETNAME. In multi mode the name of the old bucket becomes the first folder inside it. The database stores the same paths in both releases, so no SQL is needed after the copy.
| Data | 26.0 (bucket / key) | 26.1 multi (bucket / key) |
|---|---|---|
| Attachments of project 1 |
prj-1 / 2026-11/<launchUuid>/<file>
|
rp-bucket / prj-1/2026-11/<launchUuid>/<file>
|
| Integration salt |
prj-keystore / secret-integration-salt
|
rp-bucket / prj-keystore/secret-integration-salt
|
| User photos |
prj-users / <file>
|
rp-bucket / prj-users/<file>
|
| Plugins |
rp-bucket / <file>.jar
|
rp-bucket / rp-bucket/<file>.jar
|
Every 26.0 bucket becomes a folder of the same name. That includes the default bucket itself, which holds the plugins.
The database never stores a bucket. attachment.file_id holds the base64 of the relative path, for example 1/2026-11/<launchUuid>/<file>. 26.0 and 26.1 build that path the same way. Only the step from path to storage changed:
| Mode | Path in the database | 26.0 storage | 26.1 storage |
|---|---|---|---|
multi |
1/2026-11/<launchUuid>/<file> |
bucket prj-1, key 2026-11/<launchUuid>/<file>
|
bucket rp-bucket, key prj-1/2026-11/<launchUuid>/<file>
|
single |
project-data/1/2026-11/<launchUuid>/<file> |
bucket rp-bucket, same key |
the same, no change |
DATASTORE_BUCKETPREFIX is only the folder prefix. It does not have to be prj-, but it must be the value that 26.0 used.
In single mode new attachment keys are:
<DATASTORE_DEFAULTBUCKETNAME>/project-data/<projectId>/<yyyy-M>/<launchUuid>/<file>
A 26.0 install that already used RP_FEATURE_FLAGS: singleBucket keeps its layout. It needs no object copy.
The analyzer always uses one bucket. With the prefix override in x-analyzer-environment, 26.1 stores models at:
<DATASTORE_DEFAULTBUCKETNAME>/analyzer/prj-<projectId>/
The compose file does not move existing model files.
The integration salt is the one object you must not lose. If the API does not find it, it generates a new one and existing integration secrets in the database can no longer be decrypted. In multi mode it is read from <DATASTORE_BUCKETPREFIX>keystore/secret-integration-salt in the default bucket (default prj-keystore/secret-integration-salt). RP_INTEGRATION_SALT_PATH stays keystore.
Use this when 26.0.7 used DATASTORE_TYPE: filesystem. The files stay on the storage volume, mounted at /data/storage on the API and Jobs. The layout is the same, so there is no copy.
- Back up PostgreSQL and the storage volume.
docker compose up -d postgres
docker compose exec postgres pg_dump -U rpuser reportportal > backup-26.0.7.sql
docker run --rm -v reportportal_storage:/data alpine tar -czf - -C /data . > storage-backup-26.0.7.tar.gzIf the volume name differs, read it from docker volume ls. A host bind at ./data/storage is archived with tar -czf storage-backup-26.0.7.tar.gz ./data/storage.
- Save the file you are about to replace, and note any
DATASTORE_*lines you changed. On a default 26.0.7 install the only active line isDATASTORE_TYPE: filesystemin both environment blocks.
cp docker-compose.yml docker-compose.yml.26.0.7
cp .env .env.26.0.7- Check out the 26.1 compose file. Image tags come from
.env, not from*_VERSIONnames. 26.0.7 and 26.1 both readAPI_IMAGE,JOBS_IMAGE,UI_IMAGE,MIGRATIONS_IMAGE, andANALYZER_IMAGE. When a variable is unset, the compose file uses its own default (5.15.xon 26.0.7,6.0.0on 26.1). If.envpins the old images, point them at6.0.0.
git fetch origin
git checkout 26.1API_IMAGE=reportportal/service-api:6.0.0
JOBS_IMAGE=reportportal/service-jobs:6.0.0
UI_IMAGE=reportportal/service-ui:6.0.0
MIGRATIONS_IMAGE=reportportal/migrations:6.0.0
ANALYZER_IMAGE=reportportal/service-auto-analyzer:6.0.0- Confirm
x-storage. The shipped block is already the filesystem setup. Leave it as it is unless step 2 found a custom container path.
x-storage: &storage-environment
DATASTORE_TYPE: filesystem
DATASTORE_PATH: /data/storageapi and jobs still mount storage:/data/storage. If 26.0.7 used a bind mount, put that bind back on both services. DATASTORE_PATH must be the path inside those containers, not the host path. The analyzer ignores DATASTORE_PATH. Confirm x-analyzer-environment still has FILESYSTEM_DEFAULT_PATH: /data/storage/analyzer and does not set ANALYZER_BINARYSTORE_TYPE. The analyzer-storage volume is mounted at that path.
- Start 26.1.
--remove-orphansdeletes theuatandindexcontainers, which this file no longer defines.
docker compose pull
docker compose up -d --remove-orphans
docker compose logs -f migrationsWait until the migrations log reports that the migration finished, then check the stack:
docker compose pspostgres, rabbitmq, opensearch, api, jobs, ui, analyzer, and gateway are running. migrations has exited 0. uat and index are gone.
- Sign in through the API. Open
http://localhost:8080and sign in asadmin@reportportal.internalwith the password you used forsuperadmin. Open a launch from 26.0 and download an attachment. Scripts that request tokens use/api/sso/oauth/tokenwith clientui:uiman.
Use this when you run the object store yourself and keep it: external MinIO, SeaweedFS, Ceph RGW, or another S3 API (26.0 DATASTORE_TYPE: minio, or s3 with a custom endpoint). The store stays where it is. You move the settings into x-storage, and for multi you copy inside the store.
In 26.0, multi means one bucket per project (prj-<projectId>). In 26.1, multi keeps the same names but as folders inside the default bucket (rp-bucket/prj-<projectId>/…). 26.1 does not read the 26.0 buckets.
For single (RP_FEATURE_FLAGS: singleBucket), do steps 1, 2, 5, and 6. Skip the copy.
- Back up PostgreSQL and save the 26.0.7 compose file, including the
DATASTORE_*lines fromx-environmentand fromx-analyzer-environment.
docker compose exec postgres pg_dump -U rpuser reportportal > backup-26.0.7.sql
cp docker-compose.yml docker-compose.yml.26.0.7
cp .env .env.26.0.7- Put the storage settings into
x-storageon the 26.1 file. Set the*_IMAGEtags from Option A step 3. Keep the bucket name, prefix, postfix, region, and keys you used in 26.0. Write the secret asDATASTORE_SECRETKEY.
x-storage: &storage-environment
DATASTORE_TYPE: s3-compatible
DATASTORE_ENDPOINT: "http://minio:9000" # scheme and port, as in 26.0
DATASTORE_REGION: ""
DATASTORE_ACCESSKEY: ${STORAGE_ACCESSKEY-rpuser}
DATASTORE_SECRETKEY: ${STORAGE_SECRETKEY-rppassword}
DATASTORE_DEFAULTBUCKETNAME: rp-bucket
DATASTORE_BUCKETPREFIX: prj-
DATASTORE_BUCKETPOSTFIX: ""
RP_INTEGRATION_SALT_PATH: keystore
# RP_FEATURE_FLAGS: singleBucket # only if 26.0 already used itIn x-analyzer-environment, keep FILESYSTEM_DEFAULT_PATH and set the analyzer prefix. Delete ANALYZER_BINARYSTORE_TYPE. Do not set DATASTORE_TYPE or DATASTORE_DEFAULTBUCKETNAME again in that block. The analyzer takes DATASTORE_TYPE from x-storage.
x-analyzer-environment: &common-analyzer-environment
<<: *storage-environment
FILESYSTEM_DEFAULT_PATH: &analyzer-default-path /data/storage/analyzerx-environment must not contain DATASTORE_*. It only merges the anchor:
x-environment: &common-environment
# ... database and AMQP ...
<<: *storage-environment- Only for
multi: stop the services that write objects.
docker compose stop api jobs analyzer uatuat is still the 26.0 service at this point. If it is already gone, omit it.
- Only for
multi: copy inside the store. The copy reads the old buckets and writes folders into the default bucket. The old buckets stay as the rollback copy.
# macOS: brew install minio/stable/mc
# Linux: wget https://dl.min.io/client/mc/release/linux-amd64/mc && chmod +x mc
mc alias set store http://localhost:9000 <access-key> <secret-key>
mc mb --ignore-existing store/rp-bucket
# Plugins that already sit in the default bucket become rp-bucket/rp-bucket/
mc mirror --preserve --exclude "rp-bucket/*" store/rp-bucket store/rp-bucket/rp-bucket
# Every other bucket becomes a folder of the same name
mc mirror --preserve --exclude "rp-bucket/*" store/ store/rp-bucket/Use your endpoint, keys, and DATASTORE_DEFAULTBUCKETNAME if they are not the defaults above. For a store on another host, set the alias to that URL (https://s3.example.com:443 when the store requires TLS).
Check the salt before you start 26.1. The API reads it once, at start.
mc stat store/rp-bucket/prj-keystore/secret-integration-saltThe result is one folder per old bucket:
rp-bucket/
├── prj-1/
├── prj-keystore/secret-integration-salt
├── prj-users/
└── rp-bucket/
- Start 26.1.
docker compose pull
docker compose up -d --remove-orphans
docker compose logs -f migrations- For
multi, restart the API and Jobs after the copy. This is required. The API reads the integration salt once, at start, so it must be restarted to use the copied salt. Restarting Jobs is harmless.
docker compose restart api jobs
docker compose ps- Check. Wait until
apiis healthy, then confirm:
- an attachment that was uploaded on 26.0 opens
- the connection test of an existing integration works. This confirms that the old salt is in use
- the plugins page loads
- for
multi,mc stat store/rp-bucket/prj-keystore/secret-integration-saltstill shows the salt you copied - you can sign in as
admin@reportportal.internal
Use this when 26.0.7 used DATASTORE_TYPE: s3 with RP_FEATURE_FLAGS: singleBucket. The bucket and the layout stay the same, so there is no copy.
- Back up PostgreSQL and save the compose file, as in Option B step 1.
- Move the settings into
x-storage. Keep the region, the bucket name, andsingleBucket. Keep the keys you use, or leave them empty when the host role (an EC2 instance profile) provides them. The 26.0.7 example nameDATASTORE_SECRETSKEYbecomesDATASTORE_SECRETKEY.
x-storage: &storage-environment
DATASTORE_TYPE: aws-s3
DATASTORE_REGION: us-east-1
DATASTORE_ACCESSKEY: ""
DATASTORE_SECRETKEY: ""
DATASTORE_DEFAULTBUCKETNAME: my-rp-bucket
RP_FEATURE_FLAGS: singleBucketSet the analyzer prefix in x-analyzer-environment, as in Option B step 2, if the analyzer used <bucket>/analyzer in 26.0.7. Remove every DATASTORE_* line from x-environment. Set the *_IMAGE tags from Option A step 3.
- Start 26.1.
docker compose pull
docker compose up -d --remove-orphans
docker compose logs -f migrations- Check. Wait until
apiis healthy, then confirm that an attachment from 26.0 opens, and that the connection test of an existing integration works. Sign in asadmin@reportportal.internal.
- Back up PostgreSQL before you change the release. For filesystem storage, also back up the
storagevolume. Formultiobject storage, the copy only creates folders; it does not delete the original buckets, and those buckets are the rollback copy. - The database is not changed by the object copy. The 26.1 migrations change the schema when the new stack starts.
- To roll back, stop 26.1, return to the 26.0.7 compose file and
.env, restore the database, and start that file.
docker compose down
git checkout 26.0.7
cp .env.26.0.7 .env
# If you edited the 26.1 file in place instead of checking out a tag:
# cp docker-compose.yml.26.0.7 docker-compose.yml
docker compose up -d postgres
docker compose exec postgres psql -U rpuser -c "DROP DATABASE reportportal;"
docker compose exec postgres psql -U rpuser -c "CREATE DATABASE reportportal;"
docker compose exec -T postgres psql -U rpuser -d reportportal < backup-26.0.7.sql
docker compose up -dFilesystem bytes are still on the volume, because 26.1 uses the same mount. For multi, 26.0 reads the original buckets, which the copy left in place.
- Docker Compose does not copy objects and does not rewrite
attachment.file_id. The copy in Option B is the only data move. Filesystem andsingleneed none. - The analyzer looks for models at
<bucket>/analyzer/prj-<projectId>/after the upgrade. Confirm one model; the compose file does not move it. On filesystem storage the models stay on theanalyzer-storagevolume atFILESYSTEM_DEFAULT_PATH(/data/storage/analyzer).DATASTORE_PATHdoes not point at that directory. - Leaving
ANALYZER_BINARYSTORE_TYPEset makes service-auto-analyzer 6.0.0 log a deprecation warning on start. Remove the variable. The type in use isDATASTORE_TYPE. - 26.1 renames the built-in user in the database (
superadminbecomesadmin@reportportal.internal). Sign in with the new name. Accounts from LDAP or SAML are unchanged. API tokens remain valid. - Clients that called
/uat/sso/oauth/tokencall/api/sso/oauth/tokeninstead. - The plugin copy leaves the old and the new plugin jar versions side by side under
rp-bucket/rp-bucket/. The page still loads. Remove the old jars once you no longer need the 26.0 copies. -
uatorindexstill running means the stack was started without--remove-orphans. Rundocker compose up -d --remove-orphansagain.