This project is mirrored from https://gitee.com/cowcomic/pixie.git.
Pull mirroring failed .
Repository mirroring has been paused due to too many failed attempts. It can be resumed by a project maintainer.
Repository mirroring has been paused due to too many failed attempts. It can be resumed by a project maintainer.
- 01 Feb, 2022 3 commits
-
-
Vihang Mehta authored
Summary: Buildbuddy - Added more info about handling cert-mananger and nginx ingress Sourcegraph - A lot of stuff has changed upstream. I didn't manage to get GCE ingress to work, so switched to nginx ingress. Test Plan: Deployed to new devinfra cluster Reviewers: michelle, zasgar Reviewed By: michelle Signed-off-by:
Vihang Mehta <vihang@pixielabs.ai> Differential Revision: https://phab.corp.pixielabs.ai/D10603 GitOrigin-RevId: 270ad4d25d3bff6243988076df22825ef439bde9
-
Vihang Mehta authored
Summary: github changed some stuff that caused the checksum for some zip archives to change. This seemingly didn't affect `.tar.gz` files. So update `rules_jvm_external` to fix the build. See also: https://github.com/bazelbuild/rules_jvm_external/issues/665 https://twitter.com/shs96c/status/1488480089700503558 Test Plan: This diff fixes our build, which is broken due to the changes in hash sums on Github. The expected state, after applying this diff, is that builds succeed both locally and on Jenkins. Reviewers: zasgar, oazizi, jps, #third_party_approvers, michelle Reviewed By: oazizi, jps, #third_party_approvers, michelle Signed-off-by:
-
Omid Azizi authored
Summary: This reverts commit a73e00c1f6c3cdb5fd182601e446e7a2bdff1424. The commit had a compatibility issue with the gcc-based builds. Test Plan: N/A Reviewers: #stirling, yzhao Reviewed By: #stirling, yzhao Signed-off-by:
Omid Azizi <oazizi@pixielabs.ai> Differential Revision: https://phab.corp.pixielabs.ai/D10632 GitOrigin-RevId: 6b7e4b3dfd18de09c9b51e2ab08c3f7541aba6bd
-
- 31 Jan, 2022 4 commits
-
-
Nick Lanam authored
Summary: Part of removing hardcoded values all over. Test Plan: Visually compare to main wherever monospace text appears. It should look the same as before if fonts load. If fonts load slowly or fail, we should now always see the same fallback. Reviewers: michelle, vihang Reviewed By: michelle JIRA Issues: PC-1405 Signed-off-by:
Nick Lanam <nlanam@pixielabs.ai> Differential Revision: https://phab.corp.pixielabs.ai/D10625 GitOrigin-RevId: aa12af3676adda75bf343c813629ea1560746eb5
-
yzhao1012 authored
Summary: Make the HTTP2 specialization close to the original template for easier contrasting. For example, the missed CONN_TRACE could be discovered easier. Test Plan: Jenkins Reviewers: #stirling, oazizi Reviewed By: #stirling, oazizi Subscribers: oazizi Signed-off-by:
yzhao1012 <yzhao@pixielabs.ai> Differential Revision: https://phab.corp.pixielabs.ai/D10617 GitOrigin-RevId: a73e00c1f6c3cdb5fd182601e446e7a2bdff1424
-
Phillip Kuznetsov authored
Summary: We had an unnecessary argument in StartStream() and we were being too defensive about the nats channel. This diff removes those two things. Functionality should remain the same. Test Plan: Tested on skaffold. Received/sent messages. Recovered during cloud outages. Reviewers: michelle Reviewed By: michelle Signed-off-by:
Phillip Kuznetsov <pkuznetsov@pixielabs.ai> Differential Revision: https://phab.corp.pixielabs.ai/D10609 GitOrigin-RevId: 7a9f7f0dcb1d968785dabb0439def928c079f2c3
-
James Bartlett authored
Summary: There was a lot of pain making changes to bcc, because bazel would rebuild all of bcc and all of bpftrace everytime you made a change. This diff allows people to locally develop external dependencies and build their artifacts themselves. This way people can take advantage of make/cmake incremental builds while developing locally. By default, this diff doesn't change the build process at all. But if lines in `repository_locations.bzl` are uncommented, then people can use the new local behaviour for bcc and bpftrace (also if desired other repos, but they aren't setup in this diff). To achieve this, I add a bazel rule called `local_cc` that mimicks a foreign_cc rule, but instead of running make or cmake, it just copies the artifacts/includes from the specified directories. This does mean people have to setup the build environment for the external dependencies themselves (when running with the new local repositories, obviously default builds don't have this requirement), but I think people will be happy to do that in exchange for not waiting 3-5 minutes for bcc and bpftrace to rebuild. Test Plan: Skaffolded with local changes to bcc and bpftrace, and saw that the changes were in effect on the cluster. Skaffolded with the local_repositories commented out, and saw a normal build without the changes to bcc, as expected. Reviewers: #stirling, vihang, zasgar, #third_party_approvers, yzhao Reviewed By: #stirling, zasgar, #third_party_approvers, yzhao Subscribers: yzhao Signed-off-by:
James Bartlett <jamesbartlett@pixielabs.ai> Differential Revision: https://phab.corp.pixielabs.ai/D10583 GitOrigin-RevId: 6b033ef752b879511de258b51cae34a4fa546a26
-
- 28 Jan, 2022 10 commits
-
-
Nick Lanam authored
Summary: Laying a framework for UI integration testing. Actions taken: - `yarn add --dev cypress` - Added a script shortcut in `package.json` - Ran `yarn cypress:open` - After that worked, cleared out the example files - Added a single test suite for auth - With env file config and sample - Tweaked eslint config for Cypress globals Here's what the GUI looks like with the sample test showing successful access to the Live UI: {F176859} Test Plan: - `yarn install` - Prepare environment variables following the instructions in `cypress/README.md`: - Find your base URL - it's where you go when you run `yarn dev` to test your local UI. - Get your Google session cookie for that environment (it's the value of the cookie `default-session5`). - If you don't want to put the cookie on the command line later, you can copy `cypress.template.env.json` to `cypress.env.json` and put the cookie's value there. If you specify the env var on the command line AND in this file, the command line value takes priority. - `yarn dev` to start the UI. Where you access this is the `CYPRESS_BASE_URL`, port included! - `CYPRESS_BASE_URL='...' CYPRESS_GOOGLE_COOKIE='...' yarn cypress:open` to play around with the GUI. It does quite a lot of stuff! - `CYPRESS_... yarn cypress:run` to run the tests headlessly, entirely in your CLI. Reviewers: michelle, vihang, #third_party_approvers, zasgar Reviewed By: michelle, #third_party_approvers JIRA Issues: PC-1411 Signed-off-by: -
Omid Azizi authored
Summary: Expose some flags so users can reduce the perf buffer sizes. This may be useful for a future low-memory mode. Test Plan: Rely on existing tests to validate the default values for now. Reviewers: #stirling, jamesbartlett, yzhao Reviewed By: #stirling, yzhao Subscribers: yzhao Signed-off-by:
-
Omid Azizi authored
Summary: Use the new pl_aux_go_binary() rule to simplify the build of the go binaries in go_grpc_client/server. Test Plan: Existing tests. Reviewers: #stirling, yzhao Reviewed By: #stirling, yzhao Signed-off-by:
-
Omid Azizi authored
Summary: The pattern of building Go binaries in containers as Stirling test binaries is a common one. This introduces a macro to simplify the process of creating such binaries. Follow-ups will apply this pattern to some of the other Stirling test go binaries we have in our repo. Test Plan: Existing tests Reviewers: #stirling, yzhao Reviewed By: #stirling, yzhao Subscribers: yzhao Signed-off-by:
-
Nick Lanam authored
Summary: Lots of classes/styles that weren't used; just removing them. A couple of specific rules were also no-ops (due to being invalid), so they were removed too after checking that they looked correct as-is. Test Plan: Smoke test. No visible change; this was dead code. Reviewers: michelle, vihang Reviewed By: michelle JIRA Issues: PC-1409 Signed-off-by:
-
Nick Lanam authored
Summary: TSIA. For easier troubleshooting and consistency. Test Plan: N/A; only changes CSS class names. Reviewers: michelle, vihang Reviewed By: michelle JIRA Issues: PC-1412 Signed-off-by:
-
Pete Stevenson authored
Summary: For Java symbolization, we will use a JVMTI agent. This diff adds the .so lib files, for this agent, into the PEM (Pixie Edge Module) image. At this time, there are two agent libs: one to link vs. glibc, and one to link vs. musl. Test Plan: None at this time. A later diff that completes the Java symbolization effort will include a test. Reviewers: #stirling, oazizi Reviewed By: #stirling, oazizi Subscribers: oazizi JIRA Issues: PP-2982 Signed-off-by:
Pete Stevenson <jps@pixielabs.ai> Differential Revision: https://phab.corp.pixielabs.ai/D10616 GitOrigin-RevId: 3f56d28b4409e9f2821b764b244f007b88534d7d
-
Michelle Nguyen authored
Summary: We currently only build helm charts for official operator release versions. However, this makes testing the helm chart for RCs difficult (currently manually update the operator_build_release to deploy to the test charts bucket). We should just update our release build to always write rcs to our test charts. Also added a step in our operator testing checklist for testing the dev helm chart. Test Plan: Deploy an RC, check GCS to confirm the charts from been pushed to dev Reviewers: vihang, zasgar Reviewed By: vihang Signed-off-by:
Michelle Nguyen <michellenguyen@pixielabs.ai> Differential Revision: https://phab.corp.pixielabs.ai/D10614 GitOrigin-RevId: 7f2b19c633139484d1649fcca258e6465bd4c288
-
Nick Lanam authored
Summary: TSIA. Test Plan: Load Pixie (and sign in) in an incognito tab in a Chrome browser. Or any other security context where a browser blocks `sessionStorage`. Reviewers: michelle, vihang Reviewed By: vihang JIRA Issues: PP-3213 Signed-off-by:
-
Pete Stevenson authored
PerfProfiler: copy up & rename agent.so libs (will want this later when we copy them into the pem image). Summary: We need unique basenames for agent.so files, based on their libc linkage. We plan to put them into the pem image under /pl, and later copy them into /tmp in the target container mount namespace. Test Plan: Existing. Reviewers: #stirling, oazizi Reviewed By: #stirling, oazizi Subscribers: oazizi JIRA Issues: PP-2982 Signed-off-by:
-
- 27 Jan, 2022 13 commits
-
-
Omid Azizi authored
Summary: Expose the mux flag so users can manually enable Mux tracing if desired. Test Plan: N/A Reviewers: #stirling, jps Reviewed By: #stirling, jps Signed-off-by:
-
Pete Stevenson authored
PerfProfiler: add a headers only pl_cc_library to the java/agent sub-tree to expose the symbol update class. Summary: To read the Java symbol file, the Java symbolizer will need access to the 'RawSymbolUpdate' struct. Here, we enable this by adding a headers only pl_cc_library into the java/agent sub-tree. Test Plan: Existing. Reviewers: #stirling, oazizi Reviewed By: #stirling, oazizi Subscribers: oazizi JIRA Issues: PP-2982 Signed-off-by:
-
James Bartlett authored
Summary: I added a patch on our fork of BCC that prevents a bug where `perf_buffer->poll()` can end up running forever (or a very long time at least). Previously, BCC would keep reading from the perf ring buffer even after reading a full cycle of the ring buffer. This change makes its so that it will only ever read at most one ring buffer worth of data. BCC PR: https://github.com/iovisor/bcc/pull/3822 Test Plan: Tested in my local skaffold + existing BPF tests. Reviewers: #stirling, #third_party_approvers, oazizi Reviewed By: #stirling, #third_party_approvers, oazizi Subscribers: yzhao, oazizi, vihang Signed-off-by:
-
Michelle Nguyen authored
Summary: This namespace field is required and filled in by Helm Test Plan: deploy local helm chart Reviewers: vihang Reviewed By: vihang Signed-off-by:
-
Michelle Nguyen authored
Summary: Some of our developers are already developing on 16.13.2 locally, so might as well upgrade on all of our machines for consistency. Test Plan: ./scripts/run_docker.sh, check `node --version` and try to build the UI. Reviewers: nlanam, vihang, zasgar Reviewed By: nlanam, vihang Signed-off-by:
-
Pete Stevenson authored
Summary: TSIA Test Plan: Existing. Reviewers: #stirling, oazizi Reviewed By: #stirling, oazizi JIRA Issues: PP-2982 Signed-off-by:
-
yzhao1012 authored
Summary: Ran into stack limit during debugging. Since the data event is already using heap buffer, this diff just does the same for header events. The only complication is setting end_stream to false. Test Plan: Jenkins Reviewers: #stirling, oazizi Reviewed By: #stirling, oazizi Subscribers: oazizi Signed-off-by:
-
Pete Stevenson authored
Summary: This adds a wrapper for stat and for chown into our fs_wrapper. This work is motivated by a need to use both stat & chown in the Java symbolizer in the Stirling perf profiler. Test Plan: None: these are very thin wrappers. A test for these requires careful construction and possibly root access (which would upgrade the requirements of the existing fs_wrapper_test). Reviewers: #stirling, oazizi Reviewed By: #stirling, oazizi JIRA Issues: PP-2982 Signed-off-by:
-
Omid Azizi authored
Summary: Using a deterministic containerized build is required so that we can upgrade our Go toolchain without breaking tests. As a side effect, it also lets us get rid of the PL_COVERAGE ifdefs. Test Plan: Existing tests Reviewers: #stirling, jps Reviewed By: #stirling, jps Signed-off-by:
-
Vihang Mehta authored
Summary: The kubectl was missing a apply. Test Plan: N/A Reviewers: michelle, zasgar Reviewed By: michelle Signed-off-by:
-
Pete Stevenson authored
Summary: Add attach.cc into the perf-profiler java subtree. Essentially, it is a wrapper to invoke jattach which is responsible for injecting a .so lib into a running Java process as an "agent" (under the auspices of the JVMTI or Java Virtual Machine Tools Interface). Besides being this wrapper, attach.cc also copies the agent libs into the target process namespace and makes a decision about which lib (glibc or musl) to use. Finally, all of this is done in a separate process because we don't want to block the main process while the attach mechanism is invoked (jattach, and really any attach code, may need to sleep for a while waiting on Java). Test Plan: attach_test.cc Reviewers: #stirling, oazizi Reviewed By: #stirling, oazizi Subscribers: yzhao, oazizi JIRA Issues: PP-2982 Signed-off-by:
-
Omid Azizi authored
Summary: This new custom chunked transfer decoder is a replacement for the existing one by pico. This version does not mutate the input buffer on a parse failure, which is important because we may have incomplete data at the time of the parse. By doing so, we can save a copy of the data buffer which was otherwise required when using pico's parser. Test Plan: Numerous test cases added. Still need to evaluate the performance and memory characteristics of this version. Reviewers: #stirling, jamesbartlett Reviewed By: jamesbartlett Subscribers: yzhao, rcheng, jamesbartlett Signed-off-by:
-
Omid Azizi authored
Summary: This lets us control the order in which the sources come up. In particular, we want socket tracer to come up before the profiler, in case memory runs out. Test Plan: Manual Reviewers: #stirling, yzhao Reviewed By: #stirling, yzhao Subscribers: yzhao Signed-off-by:
-
- 26 Jan, 2022 8 commits
-
-
Vihang Mehta authored
Summary: Use new cluster for jobs Test Plan: Jenkins should work Reviewers: zasgar, michelle Reviewed By: michelle Signed-off-by:
-
Omid Azizi authored
Summary: Add knobs to control the overprovision factors of eBPF data structure sizes. Also reduce the default values a bit to reduce memory requirements. Test Plan: These knobs are for back-up controllability. Rely on existing tests to cover the default values. Reviewers: #stirling, jps Reviewed By: #stirling, jps Subscribers: jps Signed-off-by:
-
Pete Stevenson authored
Summary: Allows us to run skaffold and explicitly selcect a 'dbg' compile. Test Plan: None, this is a dev. tool. Reviewers: #stirling, michelle Reviewed By: michelle Signed-off-by:
-
Nick Lanam authored
Summary: Material deprecated it in v5. Technically all of `@mui/styles` was deprecated in favor of `@mui/system`, but we make an exception for `makeStyles` and `createStyles` due to speed and cleanliness. - Drops all uses of `withStyles` and `WithStyles` - Uses `makeStyles`, `createStyles`, and `styled` instead as appropriate - Removes components that were unused that happened to use `withStyles` Test Plan: Smoke test affected components: - Auth messages (any, such as signup completion or the callback page) - Tab headers in PxL/Vis editor - Marketing blurb in `/auth/signup` - Profile menu (dropdown in top right) - Admin cluster list - Admin cluster details (expand a row!) - Admin API/Deployment key pages. Click the icons too. - Live view data drawer - Deep links in script result tables Swapping between a tab from `main` and a tab from this branch should make differences obvious. There should be no style differences. Reviewers: michelle, vihang Reviewed By: vihang JIRA Issues: PC-1406 Signed-off-by:
-
Omid Azizi authored
Summary: Move map size computations to user-space. This will help as we try to add knobs for controlling data structure sizes with flags. Test Plan: No functional changes, so existing tests. Reviewers: #stirling, yzhao, jps Reviewed By: #stirling, yzhao, jps Subscribers: jps, yzhao Signed-off-by:
-
Vihang Mehta authored
Summary: After some testing, it appears that pixie works on k0s. Allow k0s users to deploy. Test Plan: bazel built the cli and used it to deploy on a k0s cluster. Added temporary logging to validate the checker matched. Reviewers: michelle, nserrino, htroisi Reviewed By: michelle Signed-off-by:
-
Vihang Mehta authored
Summary: Add configs to add GKE Workload Metrics scraping to our clusters. Test Plan: Applied the yamls to staging and testing. Reviewers: michelle, zasgar Reviewed By: michelle Signed-off-by:
-
Michelle Nguyen authored
Summary: We have a deleter job which runs as a pre-delete hook in Helm. This job is tied to a service account that is deployed by the operator. If the user uninstalls the helm chart before the operator is done deploying, the job will fail. Instead, we should deploy a role/binding for the deleter job at the time of the helm deploy. Test Plan: Build an rc helm chart for operator, deploy the helm chart. Verify that the clusterrole and service account exists. Uninstall the helm chart and ensure that deleter job runs successfully. Reviewers: vihang Reviewed By: vihang Signed-off-by:
-
- 25 Jan, 2022 2 commits
-
-
Nick Lanam authored
Summary: The cache persistence was reloading without cause when `getUserForContext` returned the already-stored value. One-line goof. Test Plan: Visit admin page. Create and delete a deploy key. Should not flicker out of existence. Reviewers: michelle, vihang Reviewed By: michelle Signed-off-by:
-
Vihang Mehta authored
Summary: `extensions/v1beta1` was removed in k8s `v1.22`. Migrate to `networking.k8s.io/v1` which has been available since k8s `v1.19`. Test Plan: Deploy to staging should work again. The prod cluster is tbd since the cluster version is post `1.19` but the nodepools are not, however I think GKE should handle that for us. Reviewers: zasgar, michelle Reviewed By: michelle Signed-off-by:
-
