Storing error values to atomic.Value may fail if they have different
dynamic types. Wrap error value in a consistent struct type to avoid
panics.
Make sure we return a nil error on success
If the shell spawns a subprocess which doesn't close the output file
descriptors, the exec.Cmd will block on Wait() (see
golang.org/issue/18874). Use an os.Pipe to provide the command with a
real file descriptor so the exec package doesn't need to do the copy
manually. This in turn may block our own reading goroutine, but we can
select on that and leave it for cleanup later.
Fixes#10788
This checks `IsComputed` prior to attempting to use the JSON
configurations. Due to a change in 0.8, the prior check for simply map
existence would always succeed even with a computed value (as designed),
but we forgot to update provisioners to not do that.
There are other provisioners that also do this but to no ill effect
currently. I've only changed Chef since we know that is an issue.
This issue doesn't affect 0.9 due to helper/schema doing this
automatically for provisioners.
The provisioner collected all inline commands into a single script which meant
only the exit code of the last command was actually checked for an error.
* provisioner/chef: Support named run-lists for Policyfiles
Add an optional argument for overriding the Chef Client's initial
run with a named run-list specified by the Policyfile. This is useful
for bootstrapping a node with a one-time setup recipe that deviates
from a policy's normal run-list.
* Update chef client cmd building per review feedback.
Fixes#10463
I'm really surprised this flew under the radar for years...
By having unique PRNGs, the SSH communicator could and would
generate identical ScriptPaths and two provisioners running in parallel
could overwrite each other and execute the same script. This would
happen because they're both seeded by the current time which could
potentially be identical if done in parallel...
Instead, we share the rand now so that the sequence is guaranteed
unique. As an extra measure of robustness, we also multiple by the PID
so that we're also protected against two processes at the same time.
Fixes#3605 and adds the functionality suggested in PR #7440.
This PR is using a different appraoch that (IMHO) feels cleaner and (even more important) adds support for Windows at the same time.
- Include new option in file provisioner. Now content or source can be
provided. Content will create a temp file and copy there the contents.
- Later that file will be used as source.
- Include test to check that changes are working correctly.
The script cleanup step added in #5577 was positioned before the
`cmd.Wait()` call to ensure the command completes. This was causing
non-deterministic failures, especially for longer running scripts.
Fixes#5699Fixes#5737
See the updated docs for more details and examples, but in short this
enables the `attributes` param from the Chef provisioner to accept a
raw JSON string.
Fixes#3074Fixes#3572
PR #3896 added support for passing keys by content, but in this same PR
all references to `path.Join()` where changed to `filepath.join()`.
There is however a significant difference between these two calls and
using the latter one now causes issues when running the Chef
provisioner on Windows (see issue #4039).
Builds on the work of #3846, shifting the Chef provisioner's
configuration options from `secret_key_path` and `validation_key_path`
over to `secret_key` and `validation_key`.
This Adds three new arguments `use_policyfile`, `policy_group` and `policy_name` to the Chef
provisioner. If `use_policyfile` == true, then the other arguments are required.
Still not a 100% fix, but that would require some more hacking in core
TF. If time permits I’ll have a look at that later on… But for now this
is a good fix to be able to close#2872
By prefixing them with `cmd /c` it will work with both `winner` and
`ssh` connection types.
This PR also reverts some bad stringer changes made in PR #2673
When surrounding the version with quotes, even no version (an empty
string) will be accepted as parameter. The install.sh script treats an
empty version string the same as no when version is set. So it will
then just use the latest available version.
Before this option (`os_type`) the provisioner would use the connection
type to determine the targeted OS. When not supplying a value for
`os_type`, it will fall back to the old behaviour, so this is full BC.
We need to decode both the Raw config and the parsed Config to make
sure all set keys are visible. Otherwise keys that will need to be
interpolated later, will be missing causing the validation to fail.
I added a debug log line in the last commit, only to find out it’s now
logging the same info twice. So removed the double entry and tweaked
the existing once.
The commit is pretty complete and has a tested/working provisioner for
both SSH and WinRM. There are a few tests, but we maybe need another
few to have better coverage. Docs are also included…
* We now return an error when you set the script_path to
C:\Windows\Temp explaining this is currently not supported
* The fix in PR #1588 is converted to the updated setup in this PR
including the unit tests
Last thing to do is add a few tests for the WinRM communicator…
The reason why the shebang is removed from these tests, is because the
shebang is only needed for SSH/Linux connections. So in the new setup
the shebang line is added in the SSH communicator instead of in the
resource provisioner itself…
This is needed as preperation for adding WinRM support. There is still
one error in the tests which needs another look, but other than that it
seems like were now ready to start working on the WinRM part…
Sander van Harmelen