There are two variables we use to control the volume of Top output,
maxRowsToDisplay, which controls how many rows are in the table, and
maxRowsToStore, which controls the size of the event index we keep in memory for
aggregating results.
Previously, we were only keeping in index maxRowsToDisplay rows, which for the
Resource Detail page was 10 (which is really small for high traffic rest-y
resource traffic - it causes rows to be deleted from the index too soon, and
then causes the data in the table to change a lot). Change this to store
maxRowsToStore rows, and also bump this to 50. This allows us to store results
for longer, and also ensures more consistent data over time.
Another fix for the appearance of the Top columns is to add fixed widths to the
metrics. This will prevent the table from wobbling from side to side.
A bunch of web UI tweaks:
- Add a small success rate chart to the metrics tables
- Improve latency formatting for seconds latencies
- Rename upstream/downstream to inbound/outbound
- Make Top table look consistent with rest of tables on page
- Fix widths of metrics column columns so that tables align
Consolidate the source and destination columns into one column,
and add a direction column (To/From) so the user knows if the
displayed resource is src/dst.
This PR is a result of a change request that was missed in PR #1613. This change removes an unnecessary calc() function in the sidebar.css
Signed-off-by: Dennis Adjei-Baah <dennis@buoyant.io>
The web client displays `Websocket [code]` on websocket close errors.
Modify the web client to render a more helpful error message to the
user. If a reason is present, render that, otherwise translate the
websocket error code into a message.
Fixes#1599
Signed-off-by: Andrew Seigner <siggy@buoyant.io>
This PR adds a breadcrumb style navigation to the Linkerd dashboard. Each "crumb" links to its corresponding page in the UI.
This PR also includes a small UI fix in the sidebar. The select box always seems to revert to the All Namespaces option whenever there is a state change on the React side. The fix ensures that the select box always displays the namespace filter if it is available and revert to All Namespaces when no namespace is selected.
fixes#1464fixes#1543fixes#1627
Signed-off-by: Dennis Adjei-Baah <dennis@buoyant.io>
`linkerd inject` was not checking its input for known sidecars and
initContainers.
Modify `linkerd inject` to check for existing sidecars and
initContainers, specifically, Linkerd, Istio, and Contour.
Part of #1516
Signed-off-by: Andrew Seigner <siggy@buoyant.io>
Fixes#1593
Add a `--hide-sources` flag to `linkerd top`. Setting this removes the source column from the output.
Signed-off-by: Alex Leong <alex@buoyant.io>
* Use Tap data on Resource Detail page to display unmeshed resources
that send traffic to the specified resource.
* Don't update neighbors on every websocket recv; this causes too much rendering.
Instead, store in internal variable and update with the api results.
This branch uses the src data from tap to discern which unmeshed resources are
sending traffic to the specified resource. We then show this resource in the
octopus graph.
Note that tap is sampled data, so it's possible for an unmeshed resource to not
show up. Also, because we won't know about the resource until it appears in the
Tap results, results could pop into the chart at any time.
linkerd only routes TCP data, but `linkerd inject` does not warn when it
injects into pods with ports set to `protocol: UDP`.
Modify `linkerd inject` to warn when injected into a pod with
`protocol: UDP`. The Linkerd sidecar will still be injected, but the
stderr output will include a warning.
Also add stderr checking on all inject unit tests.
Part of #1516.
Signed-off-by: Andrew Seigner <siggy@buoyant.io>
_.throttle setState for receiving websocket tap events to prevent continuous rerendering
Problem
We receive a lot of websocket events from the tap server. Previously, we
were processing each event as we received it, then calling setState after
processing to update the tables. Each call to setState triggered a re-render of
the whole table. We were rerendering multiplie times a second, causing the whole
page to become unresponsive.
Solution
Throttle setState for receiving websocket tap events to prevent
continuous rerendering. Store the tap events in an index outside of state, and
only update the state once every specified interval (currently 500ms).
We can now view entire namespaces with Top and the page won't crash!
To verify: Go to /top and try topping a namespace
When scrollbars are set to always be visible in a browser, we see them appear in the sidebar component of the dashboard.
This PR adds CSS that hides the scrollbar for WebKit browsers, i.e., Chrome and Safari and uses an overflow: hidden technique inspired by this solution to hide the scrollbar in Firefox.
fixes#1611
Signed-off-by: Dennis Adjei-Baah <dennis@buoyant.io>
If an input file is un-injectable, existing inject behavior is to simply
output a copy of the input.
Introduce a report, printed to stderr, that communicates the end state
of the inject command. Currently this includes checking for hostNetwork
and unsupported resources.
Malformed YAML documents will continue to cause no YAML output, and return
error code 1.
This change also modifies integration tests to handle stdout and stderr separately.
example outputs...
some pods injected, none with host networking:
```
hostNetwork: pods do not use host networking...............................[ok]
supported: at least one resource injected..................................[ok]
Summary: 4 of 8 YAML document(s) injected
deploy/emoji
deploy/voting
deploy/web
deploy/vote-bot
```
some pods injected, one host networking:
```
hostNetwork: pods do not use host networking...............................[warn] -- deploy/vote-bot uses "hostNetwork: true"
supported: at least one resource injected..................................[ok]
Summary: 3 of 8 YAML document(s) injected
deploy/emoji
deploy/voting
deploy/web
```
no pods injected:
```
hostNetwork: pods do not use host networking...............................[warn] -- deploy/emoji, deploy/voting, deploy/web, deploy/vote-bot use "hostNetwork: true"
supported: at least one resource injected..................................[warn] -- no supported objects found
Summary: 0 of 8 YAML document(s) injected
```
TODO: check for UDP and other init containers
Part of #1516
Signed-off-by: Andrew Seigner <siggy@buoyant.io>
Adds a new page that shows all namespaces in an accordion. This will replace
ServiceMesh as the default landing page.
The page will request stats for all namespaces, and then pick the first meshed
namespace that's not the linkerd namespace to auto-expand in the accordion.
This branch also updates the definition of "added to the mesh" in the frontend
to be runningPodCount > 0 && meshedPodCount > 0 (previously, it was
runningPodCount === meshedPodCount, which would count resources with no pods as
"added").
I've also moved the link to /namespaces out of the top-level sidebar and into
the Resources sub-menu.
* Use url query params for tap/top form filters
* Add comment explaining react-url-query onChange handlers
Signed-off-by: Kevin Lingerfelt <kl@buoyant.io>
Previously, WebSocket error messages would appear with the first
couple characters cut off. I've fixed this by using ws.WriteControl
instead of ws.WriteMessage to write errors, as gorilla does in
their example app.
- Use writeControl to write error messages to the client
- Stop the spinner if there is an error present
For a better UI experience, the sidebar should be able to scroll independently from the detail view. This PR allows both the sidebar and the detail view to scroll independently.
fixes#1547
Signed-off-by: Dennis Adjei-Baah <dennis@buoyant.io>
- Use an ant Select instead of Autocomplete for resource list, so that the user
can see all available tappable resources
- Fix bug where the authority autocomplete wasn't showing any options
- Adds "namespace/" as an option in the resource selection dropdown
This required some weird handling because we allow requests of the form
linkerd tap namespace/linkerd (taps namespace linkerd)
but not
linkerd tap namespace --namespace linkerd (does not work as intended,
taps every namespace)
Remove twitter, github and slack links from the sidebar.
The "Update Linkerd" menu item will still show up if there's an update.
The "Update now" button will also still show.
Increase the MaxRps on the tap server to 100 RPS.
The max RPS for tap/top was increased in for the CLI #1531, but we were
still manually setting this to 1 RPS in the Web UI and Web server.
Remove the pervasive setting of MaxRps to 1 in the web frontend and server
In #1540 I moved the version check code and was using our prefixed version of
fetch. This is unnecessary because the version check URL doesn't depend on the
dashboard URL prefix.
TLDR Don't use prefixed fetch for version check
A bunch of small items.
This branch:
- filters out un-meshed resources from the Tap and Top autocompletes
- removes an un-rendered title attribute from the sidebar menu items
- formats latency in Tap with a comma
- prevents the grafana link from showing if there are 0 pods in a deployment
When the mesh completion message calls to action it prints a CLI command to copy&paste. It's visually hard to separate message from the command snippet which is what this commit fixes.
Flipped background and font color to create a better visual distinction
Successfully ran web app test suite
Signed-off-by: Sebastian Tiedtke <sebastiantiedtke@gmail.com>
Previously, we included a version check in the server polling loop, which meant
we were hitting the version check endpoint once very 10 seconds from the
sidebar. This PR moves that check out of the loop so that we only hit it once,
upon pageload.
This PR also includes
- some whitespace fixes
- a fix for a console error we were triggering with our tests
The default value for the max-rps argument to the tap and top commands is an overly conservative 1rps. This causes the data to come in very slowly and much data to be discarded. Furthermore, because tap requests are windowed to 10 seconds, this causes long pauses between updates.
We fix this in two ways. Firstly we reduce the window size to 1s so that updates will come in at least once per second, even when the actual RPS of the data path is extremely high. Secondly, we increase the default max-rps parameter from 1 to 100. This allows tap to paint an accurate picture of the data much more quickly and sidesteps some sampling bias that happens when the max-rps is low.
In general, tap events tend to happen in bursts. For example, one request in may trigger one or more requests out. Likewise, a single upstream event may trigger several requests to the tapped pod in quick succession. Sampling bias will occur when the max-rps is less than the actual rps and when the tap event limit subdivides these event bursts (biasing towards the first few events in the burst). The greater the max-rps, the less the effects of this bias.
Fixes#1525
Signed-off-by: Alex Leong <alex@buoyant.io>
Previously, we would tap any resource's pods, regardless of whether the pods
were meshed or not. We can't actually tap non-meshed pods, so I'm adding a check
that will filter out non-meshed pods from the pods that tap watches.
Previous behaviour:
When attempting to hang a non meshed pod, it would establish
a watch on the pods, but then never return any results. In the CLI you could
just cancel it with Ctrl-C. In the web, clicking Stop would send a
WebSocket.close(1000) but wouldn't actually close the connection...
Behaviour after change :
If no pods under the specified resource are meshed, it'll
return an error of no pods being found to tap
Linkerd CLI's "look and feel" is similar to Kubernetes kubectl CLI. Linkerd's dashboard can be extended to match Kubernetes dashboard UI.
This PR serves as a starting point for this work. The new sidebar shows all resources from all namespaces on initial page load. Resources can be filtered to show only items in a given namespace. The sidebar displays authority, deployment, service and, pod resources. We may need to think about whether it is necessary to show all resources types. Some resources, i.e. authorities, contain a large cardinality of resource details and may not be very useful to a user.
fixes#1449
Signed-off-by: Dennis Adjei-Baah <dennis@buoyant.io>
Adds basic probes to the linkerd-proxy containers injected by linkerd inject.
- Currently the Readiness and Liveness probes are configured to be the same.
- I haven't supplied a periodSeconds, but the default is 10.
- I also set the initialDelaySeconds to 10, but that might be a bit high.
https://kubernetes.io/docs/tasks/configure-pod-container/configure-liveness-readiness-probes/
Includes a substantial refactor of Top.jsx to move the websocket
and top-request-aggregation code into a self-contained module
so that this code can be shared by /top and by each resource
detail page.
(This refactor also helps separate concerns in that
page; since that page also makes 10 second requests to the stat
api to populate the autocompletes in the form).
The TopModule uses the startTap prop to figure out whether it
should start a websocket connection and make a tap request
when mounted. (This is because the resource detail pages
start tapping immediately upon load, whereas /top can only
start once you've entered a query.
I've removed the spinner and the awaitingWebSocketConnection
state field because that now belongs in the top module. I think a
similar refactor of tap would be good before we re-add it.
Risha Mars
Dennis Adjei-Baah
Andrew Seigner
Kevin Lingerfelt
Alex Leong
Thomas Rampelberg
Sebastian Tiedtke