oracle / bosh-oracle-cpi-release Goto Github PK

From @dmutreja on November 1, 2017 9:21

BATs test failure

rspec ./spec/system/network_configuration_spec.rb:73

Copied from original issue: oracle/bosh-oracle-cpi#10

CPI implementation repo, bosh-oracle-cpi, is currently included in the release as git submodule. That arrangement works great for contributing to CPI only and allows this release repo to consume the CPI independently. Now that the CPI has matured (to a degree anyway!) , in practice the two repos will move forward in lock step. CPI version change will almost always result in bump in the release version.

A single repo will also be easier to mirror under cloudffoundry-incubator, similar to how other vendor repos are maintained.

We should get rid of the CPI submodule and include the sources directly in this repo.

To help ppl use your CPI on Oracle Cloud, they might benefit from an example bosh cloud-config and a discussion of the cloud_properties options

Congrats to the team for so quickly writing a CPI and releasing it!

From @dmutreja on November 1, 2017 9:20

BATs test failure

rspec ./spec/system/network_configuration_spec.rb:31

Copied from original issue: oracle/bosh-oracle-cpi#9

From @dmutreja on February 27, 2018 19:13

Encountered "Error: CPI error 'Bosh::Clouds::CloudError' with message 'Error attaching volume: [POST /volumeAttachments/][409] attachVolumeConflict &{Code:0xc4204418e0 Message:0xc420441900}' in 'attach_disk' CPI method"

• Was running BOSH 261.4 with concourse deployment on it
• Upgraded to BOSH 262 using create-env. Enabled UAA and CredHub as part of the upgrade
• Redeployed concourse (with credhub enabled in web atc job) which resulted in the above error

Logging this issue for further investigation.

Copied from original issue: oracle/bosh-oracle-cpi#23

From @dmutreja on November 1, 2017 1:49

This relates to issue #6. Make sure that the CPI configuration documentation clearly mentions the supported disk size(s), particularly the constraint that persistent disk volumes must be of size 50GiB or higher.

Copied from original issue: oracle/bosh-oracle-cpi#7

From @dmutreja on January 28, 2018 22:23

Deploying Cloud Foundry intermittently fails with

L Error: CPI error 'Bosh::Clouds::CloudError' with message '[PUT /instances/{instanceId}][429] UpdateInstance default &{Code:0xc420046de0 Message:0xc420046e00}[User-rate limit exceeded.]' in 'set_vm_metadata' CPI method

Implement exponential backoff if needed.

Copied from original issue: oracle/bosh-oracle-cpi#21

From @brunoborges on December 12, 2017 8:27

For Wercker or Travis CI

Copied from original issue: oracle/bosh-oracle-cpi#18

Ops file

Rename the variable to ad and adjust the variables file created by terraform installer repo accordingly.

From @dmutreja on December 11, 2017 20:27

create_stemcell implementation currently takes anywhere from 12-15 minutes to complete. Most of that time is spent in waiting for the OCI import image operation to complete. It uses the retry-with-delay-between-retries strategy for this wait period. The delay is currently set to two minutes, but that is not optimal for cases when the image creation actually completes in the early part of that two minute window.

E.g. if the image import takes 12 min and 10ms to complete create_stemcell will actually wait for 14 minutes.

Logging this issues to optimize that behavior. Perhaps come up a with a retry strategy implementation that uses a logarithmic delay (as opposed to constant delay time) between retries.

Copied from original issue: oracle/bosh-oracle-cpi#17

From @dmutreja on February 27, 2018 20:14

Relates to #23 The OCI error message is missing (Code & Message ptrs instead) from the error thrown by attach_disk method

Copied from original issue: oracle/bosh-oracle-cpi#24

From @dmutreja on November 1, 2017 9:18

BATs test failure

rspec ./spec/system/network_configuration_spec.rb:23

Copied from original issue: oracle/bosh-oracle-cpi#8

--- FAIL: Test_VmOpsUpdateMultipleInstancesConcurrently (449.87s)
assertions.go:34: Unexpected failure in updateInstance no consumer: "text/html"

Default CPI username in CPI job spec is 'ubuntu'. Now that we have stemcells coming out of bosh-linux-stemcell-builder, the the first user provisioned in those stemcells in 'vcap' and not ubuntu. Change the default accordingly.

Fix the terraform scripts too.

From @dmutreja on November 1, 2017 9:22

BATs test failure

rspec ./spec/system/with_release_stemcell_spec.rb:100

Copied from original issue: oracle/bosh-oracle-cpi#11