Update ARM base image to 3.1.10-bionic

# 1.0.10.3 (2020-11-18)
## Edge Agent
### Bug Fixes
* Fix vulnerability issues in ARM-based docker images [6603778](Azure@6603778)

## Edge Hub
### Bug Fixes
* Fix vulnerability issues in ARM-based docker images [6603778](Azure@6603778)

* Remove await from support bundle autofac construction (Azure#3453)

Autofac inconsistently breaks if you await before resolving something.

* adding bool.trueString to cleanup processor metrics label to fix it

Co-authored-by: Lee Fitchett <lefitche@microsoft.com>

This was a simple fix. 

~~The new e2e tests currently do not test the http configuration for mgmt and workload sockets, so there may be a follow up to add tests.~~
Since http is not officially supported for mgmt and workload sockets, no tests will be added.

Currently the `ValidateMetrics` parses [metric readme page](https://github.com/Azure/iotedge/blob/master/doc/BuiltInMetrics.md) to obtain all the metric keys. The test then deploy the system runtime modules and populate its metrics which then are compared against the parsed-document set.
There are a handful of not-so-positive metrics that are not being populated if the system modules operated on a happy path during the E2E test. These not-so-positive metrics are currently causing the `ValidateMetrics` to fail since they are not being populated.

Currently the `ValidateMetrics` parses [metric readme page](https://github.com/Azure/iotedge/blob/master/doc/BuiltInMetrics.md) to obtain all the metric keys. The test then deploy the system runtime modules and populate its metrics which then are compared against the parsed-document set.
There are a handful of not-so-positive metrics that are not being populated if the system modules operated on a happy path during the E2E test. These not-so-positive metrics are currently causing the `ValidateMetrics` to fail since they are not being populated.

Cherry-picked: Azure@f443dde

Cherry-picking from main branch

This PR fixes a bug in product code where the cleanup task in MessageStore.cs can potentially start before all of its settings are set, resulting in it taking default values for its settings.
Also re-enables the MultipleTTL test in MessageStoreTest.cs and makes it run with polling vs a delay, which will speed up the test.

…e/1.0.10 (Azure#4042)

Cherry pick Fix message priorities test (Azure#4015) from main to release/1.0.10

Fixing message priorities test. It's flaky. There's a delay in it, which is likely the issue. Changing it to be a poll and making it a bit longer should improve the test.

Cherry-pick from main branch to release/1.0.10
InMemoryDbStoreLimitsTest is flaky in the CI pipeline. I believe it is because we overuse the "sender1" name. The test keeps failing on "Amqp resource is disconnected," which I think we get because we're trying to access an AMQP resource that has previously closed. This would make sense if we are using the "sender1" name somewhere else around the same time.

So this pr just changes the name to be more specific to the test. I've run it ~8 times in the pipeline and haven't seen the failure, so it looks like it solves the problem.

Cherry-picking part of this commit from main branch to 1.0.10: 43e8151

This PR cleans up workarounds that were put in when the PlugAndPlay tests were first made.

Previously, we needed to use a preview version of IoTHub that supported PnP. Since then, they've gone GA, and we can use our normal IoTHub and normal EventHub endpoint.

Cherry-pick from main branch to 1.0.10
Previously, we implemented a workaround for end-to-end (e2e) tests because config.yaml + iotedged was default pulling version 1.0.9 because we use the 1.0 tag, and 1.0.9 was tagged as 1.0 version. This was an issue because there was validation logic in the e2e tests that checks that EA has been deployed correctly. But this validation logic didn't work if the EA versions were incompatible. EA 1.0.10 had a change that 1.0.9 didn't have that made the two versions incompatible. So we added a custom bootstrap image to the e2e tests that would start the config.yaml with a 1.0.10 version of EA that was compatible with the newer version of EA.

Once we tagged 1.0.10 with 1.0, all of these changes can no go away, since we don't need to have a bootstrap image anymore.
This PR removes the bootstrap workaround.

* added metrics for current number of connected clients, individual device connect/disconnected to/from iot edge
* updated doc/BuiltInMetrics.md
* modified failed connection metrics description to align with doc and added ignore disconnect in e2e test

…o release/1.0.10

1. Update ARM base image usage to version 1.0.5.6 (3.1.10-bionic-arm*)
2. Update Windows AMD64 base image to 3.1.10-nanoserver-1809

…o release/1.0.10

…#4105)

…o release/1.0.10

This rebases all changes from Edge on K8s public preview branch into latest release/1.0.10 branch.
Cherry-picked all commits from merge base to latest edge-on-k8s-public-preview branch into this PR, then cleaned it up.
Most C# code is in the "Kubernetes" project, but some changes affect the "Service" and "Core" projects.

Tests:

TEST:
Bring up k8s based on: https://microsoft.github.io/iotedge-k8s-doc/
only using the images from this PR's build.

[k8s] Update Helm charts to allow for correct proxy settings. (Azure#3641)
TEST:
Set ".Values.iotedged.data.no_proxy" on installation.
Confirm "no_proxy" environment var is set for iotedged and agent pods.

[k8s] Add all root CA certs to TLS connections (Azure#3616)
TEST:
Set up Edge runtime with self-signed certificate. Ensure it starts and all modules connect.

[k8s] Allow Edge on K8s modules to join HostNetwork. (Azure#3618)
TEST:
Set "createOptions.NetworkMode=Host" in a module, ensure it starts and connects to EdgeHub.

[k8s] Set resource limits and requests for Edge runtime components Azure#3666
TEST:
Start runtime with these Helm charts. Once pods are running, run
kubectl get pod -n -o yaml
For each pod.
Confirm iotedged has resource limit & request set. and qosClass: Guaranteed
Confirm edgagent has resource limit for all containers and qosClass: Guaranteed
Module will have a resource limit for proxy container, and may or may not have resource for module.

[k8s public preview] Have EdgeDeploymentOperator report status to Agent Azure#3232
TEST:

Set up a module with bad bind mount:
{"image":"mcr.microsoft.com/media/live-video-analytics:1","createOptions":{"hostConfig":{"Binds":["/home/lvaadmin/samples/output:/var/media/","//:/var/lib/azuremediaservices/"],"LogConfig":{"Config":{"max-size":"10m","max-file":"10"},"Type":""}}},"auth":null}

Add space at the end of the ACR user name in the deployment. Use a module that attempts to access the ACR.

Both setups should not crash agent. Runtime status of device on portal should show a "500" error.

[k8s] Discovered some problems with createOptions parsing. (Azure#3743)
TEST:
Set a module with Cmd, EntryPoint, WorkingDir with different capitalization and ensure these are parsed and assigned in pod.

[k8s] Environment Variables can be null or empty. (Azure#3780)
TEST:
In Agent section of Helm charts, set an empty environment variable, ensure iotedged starts and launches Agent with empty variable.

In any module's environment variable section, make an unset environment variable, and one that is all "=", ensure Pod is created with these environment variables present.

* added metrics for current number of connected clients, individual device connect/disconnected to/from iot edge

* fixed metrics description

* fixed UT ConnectionManagerTest by adding missing mock which impacted by metrics change

* Update edge-hub/src/Microsoft.Azure.Devices.Edge.Hub.Core/DeviceConnectionMetrics.cs

Co-authored-by: Venkat Yalla <veyalla@microsoft.com>

* Update edge-hub/src/Microsoft.Azure.Devices.Edge.Hub.Core/DeviceConnectionMetrics.cs

Co-authored-by: Venkat Yalla <veyalla@microsoft.com>

* Update edge-hub/src/Microsoft.Azure.Devices.Edge.Hub.Core/DeviceConnectionMetrics.cs

Co-authored-by: Venkat Yalla <veyalla@microsoft.com>

* Update doc/BuiltInMetrics.md

Co-authored-by: Venkat Yalla <veyalla@microsoft.com>

* Update doc/BuiltInMetrics.md

Co-authored-by: Venkat Yalla <veyalla@microsoft.com>

* Update doc/BuiltInMetrics.md

Co-authored-by: Venkat Yalla <veyalla@microsoft.com>

* Update doc/BuiltInMetrics.md

Co-authored-by: Venkat Yalla <veyalla@microsoft.com>

* modified to use Array.Empty

* modified failed connection metrics description to align with doc and added ignore disconnect in e2e test

* fixed typo in md file

Co-authored-by: Venkat Yalla <veyalla@microsoft.com>

…o release/1.0.10

# 1.0.10.4 (2020-12-18)
## Edge Agent
### Bug Fixes
* Fix vulnerability issues in ARM-based docker images [6603778](Azure@6603778)
* GetModuleLogs works with http endpoints [cf176be](Azure@cf176be)

## Edge Hub
### Bug Fixes
* Fix `edgehub_queue_length` metric [2861dbc](Azure@2861dbc)
* Introduce new metrics [40b2de9](Azure@40b2de9)
* Fix vulnerability issues in Linux ARM and Windows AMD docker images [73fa197](Azure@73fa197)

…o release/1.0.10

cherry-pick from master Azure#4080

…o release/1.0.10

* Remove metrics v0 (Azure#4109)

* remove metrics v0 doc (Azure#4162)

iiot usually cleans up its packages on completion, but won't be able to do this if it crashes, times out, or is canceled. Update tests to remove these packages if encountered.

…o release/1.0.10

First take on automating ARM Base Image update.

Use case scenario: 
1. tools/BaseImageUpdate/baseImage.ps1 is a helper script that a developer can use on his powershell to update the ARM base image version.  i.e. `Update-ARM-BaseImages -NewASPNetCoreVersion "2.1.23"`.
2. A developer send out a Pull request with base image changes.
3. A developer uses a pipeline `builds/misc/base-images-publish.yaml` to kick off the build & base image publication.
4. (Optional) As a good measure, once (3) is completed, a developer can run the build image pipeline to verify a successful publication.

Cherry-pick: Azure@931778a

Builds azure-iotedge and libiothsm-std RPMs for the Mariner distribution.
Builds rust 1.45 from source for Mariner as long as the official Mariner repo does not satisfy iotedge's rust>=1.40 constraint.

…o release/1.0.10

Fix Clippy "Errors" for release 1.1 branch Rust1.47.0 from Rust 1.44.1.
The fix is a subset combination of 
- Azure@d080373 (https://github.com/Azure/iotedge/pull/3268/files)
- Azure@7831fc5 (https://github.com/Azure/iotedge/pull/3696/files)

…4202)

Updating Base Images for Security Vulnerability:
https://msrc.microsoft.com/update-guide/vulnerability/CVE-2021-1723

List of Update Base Images:
- Linux AMD64 :  3.1.4-alpine3.11 --> 3.1.11-alpine3.12
- Linux ARM32 & ARM64 : 3.1.10-bionic-arm* --> 3.1.11-bionic-arm*
- Windows AMD64 : 3.1.10-nanoserver-1809 --> 3.1.11-nanoserver-1809
- Windows ARM32 : nanoserver:1809-arm32v7 (latest)

Publishing ARM Base Images:
1.0.5.7-linux-arm*

Note to Reviewer: 
Please feel free to pull the branch down to your local machine and use your favorite editor's regex to make sure the base images are updated correctly. (I wish there is an easier way to do this).

cherry-pick Azure#4166

cherry-pick Azure#4095 and Azure#4184

) (Azure#4236)

Make ARM32 Base Image Publication Pipeline Timeout to be 2hr

I took a pass though our Azure pipelines YAML files to update any Ubuntu 16.04 hosted agents to 18.04. Many of these pipelines define our Tier 1 support, and since Ubuntu 16.04 won't be supported much longer we want to pin to the OS version we support.

Note that any Docker images containing our Rust components will still be built on 16.04-based agents for now, because there's some non-trivial work to update `cross` dependencies. That's okay because we aren't officially releasing any of those images for 1.1.0.

This PR also fixes a problem with a setup step in the end-to-end tests pipeline on Windows; it was trying to use docker to do some cleanup before docker is actually installed (usually it will be installed from the previous run but you can't always count on that, e.g. for a new agent, or if someone logged onto the agent and uninstalled IoT Edge).

Introducing E2E test on EFLOW machine

 Don't use http proxy for Amqp/Mqtt over TCP

…4271) (Azure#4273)

cherry-pick Azure#4271
In some customer cases it happens that cleanup fails when it tries to update message store because value is not found in the db and throws InvalidOperationException. Catching the exception so it continues cleaning up otherwise it's not able to clean up messages for the endpoint.

This change updates our YAML pipelines **for the 1.1 release** to generate and consume bits built on/for Ubuntu 18.04 instead of 16.04. Specifically:
- Connectivity, End-to-end, Long Haul and Stress pipelines + downstream scripts: replace references to Ubuntu 16.04 artifacts with 18.04
- Edgelet Packages pipeline + downstream scripts: don't build Ubuntu 16.04 artifacts
- Remove Ubuntu 16.04 cross-compile scripts
- Various markdown files, tests, comments, and schema examples: change Ubuntu 16.04 references to 18.04

Note: this change does not update the arm-based docker containers that the K8s betas use for iotedged. They're still 16.04-based in this branch, and will be updated later.

Fix issue calling GetModuleLogs using DirectMethod with a payload that contains parameters using RF3339 DateTime format.

After commit a34e0ed, the iotedge CLI fails in Mariner because IOTEDGE_HOST was not set during compilation, so the CLI has the wrong path to iotedged's management endpoint.

To fix this, we set IOTEDGE_HOST prior to running `make install`. It was already being set prior to `make release`.

- Update ARM32/ARM64 Bionic base images to 3.1.12
- Update AMD64 (Linux) Alpine base images to 3.1.12
- Update AMD64 (Windows) Nanoserver base images to 3.1.12

Update the CHANGELOG to include the latest commits.

This change updates the default agent tag in config.yaml to 1.1.

Cherry-pick a4faab5 and update 1.1-specific Windows config.yaml.

This change upgrades the sysinfo package in the iotedge CLI and iotedged. It fixes bugs called out in Azure#2747 and Azure#3746. This change was fixed in master earlier and was supposed to by ported to the 1.1 branch, but never was.

Set env variables to direct nuget caches to the data partition instead of to the rootfs

Update Build Images and Edgelet Packages pipelines to automatically build when commits are merged to release/1.1. Also turn off all triggers on the Images Release build (currently no effect since trigger overrides are enabled in pipeline's settings).

…earer. (Azure#4542)

- Grammar and capitalization ("ID", "DPS", etc).

- Consistent wording to refer to certificates and private keys.
  ("Path of the X certificate", "Path of the private key of the X certificate").

- Clarify that all certificate and private key paths are specified as file URIs,
  with examples.

- Split DPS provisioning sections into one section per attestation method,
  and dispense with the awkward "Required if... Optional if..." prologue
  on every field.

- Clarify what `always_reprovision_on_startup` and `dynamic_reprovisioning` do
  more clearly, especially their defaults.

- Consistently document `always_reprovision_on_startup` and
  `auto_generated_ca_lifetime_days` in all config files.

…on (Azure#4451)

* modified to fail and drop message on authentication error with scope authenticator

Co-authored-by: David Lu <davilu@microsoft.com>

- Linux AMD64: `3.1.12-alpine3.12` --> `3.1.13-alpine3.13`
- Linux ARM32: `3.1.12-bionic-arm*` --> `3.1.13-bionic-arm*` (1.0.5.9-linux-arm*)
- Address CR path different : dotnet/dotnet-docker#2375

# 1.1.1 (2021-03-11)
## Edge Agent
### Bug Fixes
* Fix vulnerability issues in docker images [694bc54](Azure@694bc54)

## Edge Hub
### Bug Fixes
* Sending messages to cloud is blocked when a leaf device is disabled [756b83b](Azure@756b83b)
* Fix vulnerability issues in docker images [694bc54](Azure@694bc54)

## iotedged
### Changes
* Make comments in the config.yaml for provisioning and certificates clearer [771be81](Azure@771be81)

- Fix NugetSecurity warning for Long Haul pipeline

…4612)

Also clarify what the service uses to detect reprovisioning when
`dynamic_reprovisioning` is true.

- Update release date
- Include documentation update commit.

Adds a new job for the long haul test pipeline to run on a eflow agent.

…new dashboard queries (Azure#4695)

* add testname for 1.1 branch pipelines to show on dashboard

* follow same pattern used for testName in master branch

- Update base image for ARM*

Prepare for Release 1.1.2

Fix ValidateMetrics on ARM32

These are the current latest versions. Among other things, they contain fixes
for building with recent cmake.

API changes:

- `TEST_INITIALIZE_MEMORY_DEBUG` and `TEST_DEINITIALIZE_MEMORY_DEBUG`
  were removed from testrunnerswitch.h

  Azure/c-testrunnerswitcher@0711280

- `TOSTRING` macro was removed from ctest_macros.h in favor of
  `MU_TOSTRING` from macro-utils

  Azure/ctest@6d7a34c

- `SECURE_DEVICE_TYPE`, `UINT` and `XDA_HANDLE` are not defined anywhere.

- `REGISTER_UMOCK_ALIAS_TYPE` now requires the aliasing type to be
  the same size as the aliased type.

  Azure/umock-c@2ab7102

- On Windows, `interface` ends up being `#define`d to `struct`
  by Windows headers, which causes compile errors when it's used as
  a variable name. Rename those variables to `iface`.

Other changes:

- Fix `edgelet-utils` to build with newer Rust since `tools/check_submodules`
  depends on it but does not have its own `rust-toolchain` file.

- Fix `edgelet-kube` tests to use single-thread tokio runtime, because
  the multi-threaded runtime cannot be used in parallel correctly.

- Make some buffers used with `sprintf` larger because newer gcc warns that
  they might not be large enough for the string written to them.

…eys to disk. (Azure#4854)

schannel and native-tls require a PKCS#12 blob of the device ID certificate
to use it as a TLS client cert. Before this change, the blob was created
using openssl even on Windows. However, with such a blob, schannel creates
the private key on disk under the user's persisted keys directory.
Since iotedged runs as LocalSystem, this was
C:\ProgramData\Microsoft\Crypto\RSA\S-1-5-18

It's empirically known that schannel doesn't create the disk file if
the private key is marked as "exportable". Since "exportable" is
a Windows-specific attribute, openssl doesn't have any way to set it.
So this commit uses winapi to create the PKCS#12 blob instead.

Other changes:

- Revert two uses of single-threaded tokio runtime back to multi-threaded.
  deb2753 accidentally changed these while
  changing the tests.

Removes all containers after a device reprovision. This is done because identities for containers might no longer be valid after the reprovision. Edgelet will refresh the identities on its next run.

Resolves Azure#4458

A recent update to allow libiothsm to continue to build (deb2753) broke the Debian 8 builds in our Edgelet Packages pipeline (`builds/misc/packages.yaml`), but Debian 8 is old and unsupported so we can remove it.

- serde-yaml recent version 0.8 has fixes for known recursion issues in 0.7 as per our CI system (Reference to issue - dtolnay/serde-yaml@41d5823)
- Removed unused Cargo references to serde-yaml

)

Cherry pick Azure#3507

- Moving Azure Monitor source code to our repository.
- Integrate into appropriate pipelines

Introducing new option to trasnsform/flatten the Metrics prior to publish as IoT Messages.
**Publishing with transform option OFF:**
[{
    "TimeGeneratedUtc": "<time generated>",
    "Name": "<prometheus metric name>",
    "Value": <decimal value>,
    "Label": {
        "<label name>": "<label value>"
    }
}, {
    "TimeGeneratedUtc": "2020-07-28T20:00:43.2770247Z",
    "Name": "docker_container_disk_write_bytes",
    "Value": 0.0,
    "Label": {
        "name": "AzureMonitorForIotEdgeModule"
    }
}]

**Publishing with transform option ON:**
{
  "TimeGeneratedUtc": "<time generated>",
  "edge_device": "<device bame>",
  "instance_number": "<instance number>",
  "<prometheus metric name>": {
    "value": <decimal value>,
    "from": "<value of: 'from', 'module_name', or 'id' if exists>",
    "from_route_output": "<value of: 'from_route_output', 'route_output', or 'source' if exists>",
    "to": "<value of: 'to', 'to_route_input', or 'target' if exists>",
    "priority": <integer value of 'priority' if exists>,
    "quantile": <decimal value of 'quantile' if exists>,
    "command": "<value of 'command' if exists>",
    "disk_name": "<value of 'disk_name' if exists>"
  },
  "edgeAgent_used_memory_bytes": {
    "value": 54992896.0,
    "from": "edgeAgent"
  },
  "edgeAgent_used_memory_bytes": {
    "value": 131170304.0,
    "from": "edgeHub"
  }
}

1. Reinstitute LongHaul IotHub for both SingleNode for testing
2. Consolidate KeyVault variables so the same type of tests are sharing the same IoTHub.

Remark: With the keyVault variable consolidation (2), a developer can now inject his own keyVault into the pipeline to have any of the test pipeline connect upstream to his/her own IoTHub. The two required "secrets" in the developer's keyVaults are `IotHub-ConnStr` (IoTHub Connection String) and `IotHub-EventHubConnStr` (IoTHub's built-in EventHub Endpoint Connection String)

…4950)

The metrics collector built by the azure monitor team copied some test-specific logic from our metrics collector used for testing. I am removing it as we don't intend to provide this feature for obsv1.

…zure#4955)

This PR contains a simple E2E smoke test of the customer facing Metrics Collector. @pmzara added this E2E test then I did some cleanup and integrated it into the E2E pipeline.

Updating product info and message id

Since we can now send to Central, renaming `HubResourceId` to just `ResourceId`

…4943) (Azure#4967)

This is a cherry-pick of Azure#4943

This adds a subset of the code from https://github.com/Azure/azure-iot-sdk-csharp for SharedAccessSignature directly into our repo. This will also need to be cherry-picked into 1.1 and 1.2.

Todo:
- [x] Create UTs (UT=unit test) around this code, as there seem to be some tests that use parts of SharedAccessSignature but nothing completely dedicated
- [x] See if more code can be trimmed out that is not being used, just removed things without any references for first pass
- [x] if only using one supported .net then can maybe remove some other code here

Adding images release and publish pipelines for the metrics collector. Approach was taken from the implementation we have in `master` branch for the api proxy.

…ept tcp level connection) (Azure#4984)

cherry pick of Azure#4914

… provided (Azure#4987)

Cherry-picked: Azure@2b650a8

TL;DR - This PR chances the behavior of `GetModuleLogs` method to better facilitate the portal experience. 

Previous, when the `GetModuleLogs` is triggered from the azure portal with `tail + since + until`, all the 3 options are passed to Docker API management to fetch the log. The Docker API does the following: 
1. Fetch the log of the module
2. Apply `tail` to the log
3. Apply the `since` and `until` to the log
This results in an empty log most of the time which is not a desirable behavior. 

This PR changes the `GetModuleLogs` when the `tail + since + until` are specified together by doing the following:
1. Fetch Docker API with `since + until` log option
2. `tail` the returning log using LogProcessor (post-Akka).
This allows a customer to get a tail of log within a given time frame. 

Note: This PR doesn't change the behavior of any other option combinations (i.e. `tail + since`, `tail`, `since`, etc). These combination are still managed by Docker API.

Note2: I set RequestHandle timeout to be 30s instead of 600s by default

Cherry-pick: Azure@37c661b

- Allow Management socket to accept `timestamps` option to be pass through docker socket API.
- Update the documents for the latest API version
- The latest API is being manually added since there is a gigantic out-of-sync between the template interface `managementVersion_2020_07_07.yaml` and `/edgelet/management` generated code.

…Azure#4990)

cherry-pick: Azure#4952

Update Base Images for Security Patch
- ARM32/64 🠘 3.1.15-bionic-arm32v7
- AMD64 🠘 3.1.15-alpine3.13
- Windows64🠘 3.1.15-nanoserver-1809

Prior to this change, the Azure DevOps pipelines described by YAML files in this repo ran on either (1) Microsoft-hosted Linux and Windows agents or (2) custom-built agents hosted and maintained by the product team. The Microsoft-hosted agents are great because someone else manages their maintenance, but they're also shared among many teams so availability can be a problem. The custom-built agents have better availability but require more care and attention.

To get better availability and control with a reduced maintenance burden, this change converts the YAML pipelines to refer to 1ES-hosted agents (1ES is Microsoft's internal engineering system). These agents are defined by the product team but are maintained by 1ES.

This change only impacts three pipelines: Build Images (images.yaml), Edgelet Packages (packages.yaml), and End-to-end tests (e2e.yaml). And there are a few exceptions: Raspberry Pi OS, CentOS-7, Windows EFLOW, and proxy jobs have not been converted to 1ES either because they're not yet supported by 1ES or they simply require more work. Other pipelines will be converted in future PRs.

Other details:
- This change expands the platforms and environments that run the end-to-end tests. Ubuntu 20.04 is now covered, and both Ubuntu versions run against Docker (supported only for development scenarios) and a Microsoft-built fork of the Moby project (officially supported for production scenarios). Also, the tests run against Windows 10 Enterprise and Windows Server 2019.
- The artifact overrides task was removed from e2e-run.yaml because specifying artifact builds via the `az.pipeline.images.buildId` and `az.pipeline.packages.buildId` variables is obsolete. You can now use the Azure Pipelines "resources" feature to accomplish the same thing.
- The .NET Core 2.1 & 3.1 install tasks were removed from images.yaml because they're installed when the 1ES-hosted image is built; they don't need to be installed at runtime.
- A `docker system prune` task was removed from images.yaml because the new 1ES-hosted images are stateless, and therefore don't require cleanup between runs.
- A bug was fixed in the cargo install script wherein it didn't set up the user's PATH. This was likely hidden by the fact that the pipeline ran on stateful agents, where cargo had already been installed. Cargo install will likely become a build-time action in the future, so this task will be removed altogether.
- A bug was fixed in the openssl library build task, which again probably went undetected because it only ran on stateful agents.
- The end-to-end tests now detect a couple more cases where `iotedge list` fails because the daemon is still starting up. In such cases the tests will retry the commands rather than failing.
- Process.RunAsync was updated to stream stdout in real time rather than waiting until the call ends to return output (and potentially discarding output in some failure cases). This improves logging, but required forking the open source RunProcessAsTask project into the source code rather than taking a NuGet dependency. The changes could potentially be contributed back instead, but the project seems not to have been updated for some time. Note that a few calls to Process.RunAsync still request the old behavior (via `logVerbose: false`) because they're logging the output to file and therefore don't want to repeat it to stdout.

…Azure#4942)

Upload a support bundle from the portal, using UploadSupportBundle, causes the edge device to stop responding to iotedge list and other commands executed from the cli; in addition, the portal returns a 500 error when the user tries to deploy a new module to the device.