Scope Versions Save

Release Notes

Release 1.13.2

Mostly updates to dependencies, plus a couple of small improvements.

Thanks to everyone who contributed to this release: @gaby, @Kielek, @knrt10

Enhancements

• Set hostname to Kubernetes node name #3827
• Stop rendering if Context is cancelled #3801

Bug fixes

• Only mount plugins directory if it exists #3825
• Multitenant billing: cope with spy-interval set longer than publish-interval #3796
• Multitenant query: don't drop topologies so much #3791
• Multitenant: scan container command-lines as well as process #3789

Dependencies updates

• Update alpine base image to 3.13 #3838
• Update build tools including Go 1.16.2 #3833, #3797, #3821
• update lodash to 4.17.20 (CVE-2020-8203) #3831
• update dot-prop, webpack and terser-webpack-plugin #3816
• bump http-proxy from 1.16.2 to 1.18.1 in /client #3819
• bump elliptic from 6.4.0 to 6.5.3 in /client #3814
• bump lodash from 4.17.15 to 4.17.19 in /client #3812
• Update to most recent weaveworks/ui-components #3795
• update JS kind-of to 6.0.3 #3785

This release fixes a bug on 'kubernetes describe' operations, where the probe would spin re-opening a connection to the UI again and again after the operation had finished.

Also removes some obsolete code which connected to the unsecured local kubelet port in Kubernetes; please update your configuration to a single probe to talk to Kubernetes for the whole cluster if you haven't already. Thanks to @CiMaol for this contribution.

Bug fixes

• Stop 'kubernetes describe' operations spinning #3784
• Probe: skip publishing empty reports when publish rate is higher than collection rate #3774

Improvements

• Probe no longer talks to local kubelet #3754
• Track rounding error in multitenant billing calculation #3779

Performance

• Multitenant: merge incoming reports in collector, to save IO and query time #3780 #3781 #3782

Dependencies

• update html-webpack-plugin to most recent stable #3776
• downgrade fluent-logger-golang library used in multitenant mode #3772

Release 1.13.0

This release brings a few bug-fixes and number of performance improvements, particularly in reducing the data sent when there are many socket connections between two endpoints.

The bump in version number reflects a change in wire protocol for this change in endpoints data, and also a change in the way active controls are encoded.

Thanks to everyone who contributed to this release: @DarthSett, @sarataha, @slalwani97 and @qiell.

Bug fixes

• Remove trailing zeros in large numbers in UI #3760
• kubernetes: display pod status as "terminating" where appropriate #3729
• kubernetes: detect more 'pause' containers #3743
• Improve calculation of usage in multitenant code #3751 #3753

Performance improvements

• Elide many connections from/to the same endpoints #3709
• Remove two specialised data structures; unify with other node data #3714 #3748
• Simplify some renderers to improve performance #3747
• Slow down DNS poll interval to reduce network activity #3758

Minor improvements

• Add "user-agent" header to http calls from Scope probe #3720
• Set timestamp and window on each report #3752
• Add tracing for pipe operations #3745

Dependencies updates

• Convert to Go modules #3742
• Update to Go 1.13.9 #3766
• Go: update weaveworks, prometheus, protobuf, jaeger and aws dependencies #3745 #3756
• JavaScript: update babel, jest, webpack and other dependencies; dedupe yarn.lock #3733 #3755 #3757 #3763
• Update Alpine base image to 3.11.3 #3768

Release 1.12.0

This release is tagged 1.12 to signify a breaking change: it supports Kubernetes 'v1' object types that are needed for Kubernetes 1.16 and drops support for obsolete 'v1beta' types. #3691

Also the serialisation format changes: DNS data was accidentally renamed 'nodes' in release 1.11.6, and this release changes it back to 'DNS'. #3713

Thanks to everyone who contributed to this release: @bensooraj, @chandankumar4, @imazik, @oleggator and @qiell.

Minor improvements

• Allow user to disable plugins via command-line flag. #3703
• In the UI, replace JSON.stringify with json-stable-stringify #3701

Bugs and security fixes

• fix: report http error if /api call fails #3702
• Fix a rare crash in the ebpf connection tracker by feeding initial connections synchronously on restart. #3712
• Fix typos in debugging format strings #3695

Performance improvements

• handle IP addresses in binary rather than strings #3696
• In the multitenant app, save IO by keeping rapid-update data outside of the persistent store. #3716

Dependencies updates

• Update Go version to 1.13.0 #3692 #3698
• Update google/gopacket library #3606
• Update NodeJS to 8.12.0 and various javascript libraries #3685 #3690 #3719 #3726

Build and CI improvements:

• Run UI build container as current user to avoid files being owned by root. #3635
• Replace SASS files with CSS and JavaScript #3700
• Remove obsolete -e flag from docker login in CI #3708
• Fix favicon.ico in UI dev mode #3705
• Refactor report reading to make the code simpler #3687
• Don't import fonts when Scope UI is embedded. #3704

Release 1.11.6

This is largely a performance improvement release: the biggest change is that the probe now publishes full reports one in three times; the rest are deltas which are much smaller hence use less CPU and memory in the app. #3677

Also a new debugging summary function in the app, exposed via http #3686

Some other small performance improvements:

• perf(probe): reduce copying of nodes #3679
• perf(probe): add 'omitempty' tag to Topology.Nodes #3678
• perf(probe): update netlink library to bring in performance improvements #3681
• perf(multitenant): quantise report cache in query side of aws-collector #3671

Other changes:

• add tracing spans for rendering to the UI via websocket #3682
• update some javascript dependencies #3664
• upgrade JavaScript ui-components #3670 #3673
• fix(test-flake): poll for result in TestRegistryDelete() to avoid race #3688
• refactor: remove old unnecessary controls code #3680
• stop render package depending on probe #3675
• remove some unused string constants #3674

Release 1.11.5

A few small improvements:

• Expose probe metrics to Prometheus, if an http-listen address is given #3600
• Reduce probe resource usage caused by occasional leaks in probe endpoint reporter #3661
• Update 'eslint' JavaScript checking tool, and resolve warnings #3643

Release 1.11.4

This release contains a few fixes, one of which should improve resource usage on hosts that have a lot of TCP connections.

• Improve eBPF connection tracker to reduce the number of times it restarts and falls back to a less efficient mechanism. #3653
• Add reporter name to probe error logs. #3363 Thanks to @princerachit
• Defer metrics registration until we need it #3605
• Remove unused metric SpyDuration #3646
• Remove quay.io from release script #3657

This is a bugfix release, which should improve some cases where reports get bigger and bigger over time due to Scope not seeing connections get closed.

• Report the error and restart when something goes wrong in connection tracking #3648

Minor update:

• Updates to JavaScript dependencies, where security issues were reported: #3633
• Another fix for Scope not noticing when a container has been destroyed: #3627