v271.11.0

🔗 Improvements

• Merges in PR #2320. This PR adds BPM pre-stop scripts that send SIGINT to the main postgres database process prior to using bpm to terminate the process. See the PR for more details on exactly what changed and why.

🛠️ Bugfix

• YAML files that end with three newlines will no longer have ... emitted at the end of this file. This resolves an incompatibility introduced in v271.10.0 for consumers of these files that use their contents as YAML fragments to be inserted into other files.

v271.10.0

🔗 Improvements

• Updates Ruby version used by the Director to Ruby 3.0.
• Adds parameters that permit operators to change the preferred blobstore TLS protocols and ciphers, as well as controlling whether the server's cipher list or the client's is preferred. The default TLS version used is now 1.2. These changes should make things easier for customers whose Directors have particular TLS configuration needs. The parameters and their default values can be seen in this commit: b55c6b8
• Introduces changes that make it much easier to diagnose CPI hangs. The Director now spawns CPI children using popen3, which permits the Director to get CPI output as it is created, rather than getting it all when the CPI terminates. (PR #2319)
• Extends the DataDog Health Monitor plugin to include custom tags when creating DataDog events for BOSH HM alerts. (PR #2306)

🛠️ Bugfix

• bosh cck now always finds the correct stemcell in multi-CPI installation. Resolves GitHub issue #2287. (PR #2317)
• Restricts new IAAS-native persistent disk resize operation only operations that increase size of the disk. When shrinking persistent disks, we fall back to the classic "create a new persistent disk, move the data over, and remount the new disk" strategy. NOTE: No known CPI currently makes use of this feature, so this should not affect any released code. (PR #2318)

⚠️ Known Issues

• This release of the Director causes YAML files that end with three or more newlines to have the YAML end-of-document token added to them. (That is the ... token.) This breaks consumers of these files that were previously treating them as YAML fragments that could be copied and pasted into the middle of other YAML documents.
  • NOTE This issue is resolved in Director version 271.11.0.

v271.9.0

🔗 Improvements

• Speed up Postgres database creation on initial director deploy 3eaada6
• Reduce ignorable errors in the database log 9a2639b and 7f7c90b
• Added support for user_domain_name and project_domain_name on OpenStack d60fc2c, Thanks @andinod

🛠️ Bugfix

• Fix Health Monitor error messages when BOSH director is down 0867f33
• Attach-disk now works when used in combination with multi-CPI #2303 and 12d6435, Thanks @Justin-W
• Updated puma gem to address CVE-2021-29509

v271.8.0

🔗 Improvements

• bump nginx
• clean up unhelpful blobstore logging. Thanks @rkoster #2279

🛠️ Bug fix

• Failed deploys on network migrations no longer orphan ip addresses 6239748

v271.7.0

🔗 feature

• Added a new postgres job that is using postgres 13.2. If you are using the postgres-10 job, you should be able to switch directly over and your data will be migrated. Your existing /var/vcap/store/postgres-10 directory will not be removed. Once you have successfully upgrade and performed a backup, you should remove the old directory. If you still have an old /var/vcap/store/postgres-9 data directory from the previous postgres-9 to postgres-10 upgrade, you'll need to remove it before switching to the new postgres job as we use it to ensure you are not trying to upgrade directly from postgres-9. bbr v1.18.7 and above have postgres 13 support. Credhub support is pending see pr
• Blobstore job blobstore.max_upload_size default increased to 10gb. This will allow running bosh logs more easily when there are a large amount of logs on a VM.

🔗 fix

• Extra metadata has been removed from director metrics. This should help reduce some noise from our metrics. #2279
• Fix scheduled task cleanups should better handle large numbers of tasks. Previously the tasks were all loaded into memory to destroy them, which could cause the job to crash if there were too many tasks to clean up at once.

v271.6.0

🔗 fix

• The health-monitor job was wrapped with bpm in v271.3.0. This caused problems for some health-monitor plugins that were trying to write to their own folder in /var/vcap/sys/log. We've updated the bpm config to allow this now.
  237456e

v271.5.0

🔗 fix

• The director job was not correctly loading the mysql2 gem when the director package was compiled and run on separate VMs. Fixes the issue seen in #2295.
  5a9b68c

v271.4.0

🚨 Warning 💥: This release contains an issue when using an external mysql or mariadb database for your director. This has been fixed in v271.5.0

🔗 fix

• s3_signature_version was not being passed to the s3 blobstore client. This did mostly work previously because the blobstore client would automatically default to v4 signature version for AWS, and v2 for non AWS endpoints. But was failing if you had a non-AWS s3 compatible blobstore that required a v4 signature version.
  bb933db

v271.3.0

🚨 Warning 💥: This release contains an issue when using an external mysql or mariadb database for your director. This has been fixed in v271.5.0

🔗 dependency bumps

• Bump mariadb connector to 3.1.11

🔗 feature

• BPM wraps the health monitor process which removes a bug where you could have multiple health monitor instances
• Improves director pre-start time by stopping doing a recursive chown #2270
• Allows changing of the root domain for dns #2246

v271.2.0

🔗 fix

• Remove stale postgres pid which helps avoid lock issues on unclean restarts fa03081