| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
|
|
|
|
stable/gambia
|
|
- Improve usability when deploying or deleting various
Clover components using cloverctl
- Add yaml for each of the clover-system services
including controller, collector, spark, redis, cassandra,
clovisor and jmeter separated by k8s resource and available
under a single yaml directory
- Deploy/delete yaml for services in various subsets including:
- visibility(controller, collector, spark, redis, cassandra)
- datastore(redis, cassandra)
- validation(jmeter master/slaves(4))
- clovisor (individually)
- collector (individually)
- controller (individually)
Ex. cloverctl create system controller
Ex. cloverctl create system visibility
- Ability to add nodeport for controller separately
(cloverctl create system controller nodeport)
- Use native client-go methods to create/delete all k8s resources
used by Clover including:
- pods, deployments, statefulsets, daemonsets, services
- serviceaccounts, clusterrolebindings
- Above allows yaml override of values such as image value specified
in deployment or daemonset. This is broken down by:
- tag (ex. latest, opnfv-7.0.0) or repo'(ex. opnfv, localhost:5000)
that can be added with CLI flags as shown below:
(cloverctl create system visibility -t latest -r localhost:5000)
- defaults to -r opnfv -t latest
- Creates/deletes clover-system namespace
- Also added ability to create/delete lb service for controller
external access when running on environment such as GKE
(cloverctl <create/delete> system controller lb)
Change-Id: I2a3c6c80035d4663fa38368b3ff13e9a14090a47
Signed-off-by: earrage <eddie.arrage@huawei.com>
|
|
- Formalize use of cloverctl config file using viper
- Default name of config file is '.cloverctl.yaml'
- Look for file in home directory or relative to
cloverctl executable
- Single value in config is for 'ControllerIP: <IP Address>'
- Still try and retrieve port value of clover-controller nodeport
automatically
- Still try and retrieve IP address of clover-controller LB if
GKE environment is being used and override config file value if found
in current kube context
- Improve error messaging if controller connect fails
- Clean up the CLI help output
- Modify jmeter 'get testresult' command to avoid the use
of flags and use arguments instead
Change-Id: Id95d1b3d1079ac84996681db62da1f9b6b8a88b5
Signed-off-by: earrage <eddie.arrage@huawei.com>
|
|
Change-Id: Ied0ef2ccd4eb95cace68826a1494e82f2c18d95c
Signed-off-by: Stephen Wong <stephen.kf.wong@gmail.com>
|
|
- Change tag from local repo to opnfv for clover-spark
container, which spark-submit references
- Move update of configured visibility services into
loop
Change-Id: I9d09a851977c6006ea2cbba45d7a28be6a4660b9
Signed-off-by: earrage <eddie.arrage@huawei.com>
|
|
- Add user-agent field to jmeter config yaml and jmeter jmx
template that can be configured per URL in testplan
- Add duration time that can be used in combination with
ramp time. Duration is the total time of the test while
ramp time governs how much time is allotted to establish
all configured requests/connections
- Explicitly define clover-jmeter-master service name with
default namespace for clover-controller in clover-system
namespace to reach
- Add missing request_counts.html file
Change-Id: I5e2d388db0b73ea6c28482c94c33cc2a01f1b52e
Signed-off-by: earrage <eddie.arrage@huawei.com>
|
|
|
|
|
|
|
|
|
|
Part of Clover's initial task (during project proposal) is to explore
cloud native network tracing tools. Clover the project proposes
Clovisor as a way for Clover to use the IOvisor
(https://www.iovisor.org/) to perform network tracing. This is the first
commit of this module
Please note that the BPF code used in this commit were only tested with
Linux kernel 4.14+ --- hence, if ubuntu is used, and one prefers LTS,
the requirement is 18.04+
A different patch will track the redis and k8s API server watcher code
Also, a design doc will be added in the future (TBD)
Signed-off-by: Stephen Wong <stephen.kf.wong@gmail.com>
Change-Id: I30d9f9d474b8703097c470d39628e86bc788f9b6
|
|
- Add foundation, jquery and jquery-ui open source js/css
packages for skin, layout and AJAX
- Add plotly and candela js open source graphing packages
- Build core visibility view and provide toplevel templates,
headers, menu navigation, etc.
- Update collector/visibility APIs to align to latest Spark
analytics and UI requirements
- Refresh page elements at 5 second interval.
- Visibility controls to stop continual page refresh and clear
visibility.
- Multi-select and update service set to analyze.
- System count updates for tracing, spans, metrics schemas.
- Min/max/avg response times per service (aggregate) table.
- Graphs for: per service request counts, user-agent percentages,
per service/URL counts, per URL/status code counts, per URL counts
- Tabs for distinct URLs, status codes, user-agents, upstream cluster,
operation names, nodeids (service names)
- Table for metrics counters specified in visibility configuration.
- Metrics over time scatter plots for tracked metrics.
Change-Id: Ie3ea12fd32e0274ac52f78becc4a56c4c4134660
Signed-off-by: earrage <eddie.arrage@huawei.com>
|
|
- Docker container builds are failing due to expired links
for the jmeter binary for jmeter-master and jmeter-slave
- Updated link provided
Change-Id: Icb364bc234b118e0c00359a04b51461142f832af
Signed-off-by: earrage <eddie.arrage@huawei.com>
|
|
- Add Apache Spark 2.3 with native Kubernetes support.
- Runs self contained within K8s cluster in clover-system
namespace. One container (clover-spark) includes Clover Spark
JAR artifact. This container interacts with the
K8s API to spawn a spark-driver pod. This pod in turn spawns executor
pods to execute Spark jobs.
- Currently JAR is included in source for convenience and must be
built with sbt (install sbt and execute sbt package)
- Includes JAR from DataStax to provide Cassandra connector to analyze
Cassandra schemas as RDDs (Resilient Distributed Dataset).
- Includes Redis interface JAR to write analyzed data back to visibility
(UI, CLI or API).
- Second container (clover-spark-submit) submits Spark jobs
continuously to allow Spark to be operated entirely within the cluster.
- Two Spark jobs (CloverSlow, CloverFast) allows some analytics to be
provided in real-time and other analytics to be provided over longer
horizons.
- Each Spark job spawns two executor pods.
- Includes yaml manifest to deploy clover-spark-submit with the
necessary RBAC permissions to interact with the K8s API.
- Data analyzed includes tracing and metrics schemas obtained by
clover-collector and written to Cassandra.
- Docker builds of clover-spark and clover-spark-submit are provided
and will be pushed as OPNFV DockerHub images in a separate patch.
Change-Id: I2e92c41fd75d4ebba948c0f8cb60face57005e50
Signed-off-by: earrage <eddie.arrage@huawei.com>
|
|
This work adds testcase for validating Istio installation
Change-Id: Ie895b27eb4b3fbec23cf9519699f5fe37236153c
Signed-off-by: JingLu5 <lvjing5@huawei.com>
|
|
|
|
|
|
|
|
|
|
Else it raises exceptions in __init__().
Change-Id: I4af1f59fbb9a94da1df132e750525fcdb5d0917b
Signed-off-by: Cédric Ollivier <cedric.ollivier@orange.com>
|
|
- Modify get visibility to retrieve config and stats
- Add visibility REST API in controller to clear, set, and
get from redis
- Add example yaml to set visibility (service list by name,
metric suffixes/prefixes, and custom metrics) from CLI
- Modify example yaml to start visibility (collector) for
Istio 1.0 from CLI
Change-Id: I43304ff6d70bb4b817b345b9c383ce3621fb06c7
Signed-off-by: earrage <eddie.arrage@huawei.com>
|
|
- Supply manifest for creation PV
- Add the spinnaker as a service documentation
- set the ci.jenkins.enabled to false
Change-Id: I66d91aa133085c6c5910913da4303c05fcba9800
Signed-off-by: wutianwei <wutianwei1@huawei.com>
|
|
- Modify nginx server template to use clover-controller
in clover-system namespace for file upload
- Add ability to set nginx server paths and move uploaded files
in clover-controller and set/get file upload metadata
- Add cloverctl commands to set nginx server and lb configurations
- Add example yaml files to configure nginx server and
lb v1/v2 from cloverctl
- Modify service definition for http-lb versions to distinguish
the two versions for GRPC messaging from clover-controller in SDC
deployment yaml
Change-Id: I5c6866c5ff3de358939c58ea8c0bde64a69c6eca
Signed-off-by: earrage <eddie.arrage@huawei.com>
|
|
- Modify deployment namespace to clover-system and account
for cassandra moving to the clover-system namespace
- Increase k8s compute resource assigned to cassandra to deal
with performance issues
- Add additional fields (user-agent, request/response size,
status codes) to span schema definition and modify primary keys
- Improve exception handling to prevent collect process from
crashing
- Minor changes to support tracing/monitoring with Istio 1.0
- Inhibit logging for debug messages
- Increase time back and number of traces to fetch in
each sampling interval to deal with Jaeger REST interface
returning trace data out of order under load
(tested to 300 conn/sec; 12K connections currently)
- Move trace insert into batch mode to cassandra
- Read visibility services to analyze from redis rather than
defaults (cloverctl, UI or clover-controller REST will set)
- Remove local directory copies in docker build, as image is
based on base clover container
Change-Id: Ibae98ef5057e52a6eeddd9ebbcfaeb644caec36c
Signed-off-by: earrage <eddie.arrage@huawei.com>
|
|
JIRA: CLOVER-52
1. Add mainfest to install the spinnaker in kubernetes cluster
2. after using mainfest to install spinnaker,
we can interacte with the halyard daemon with its REST API
and we can add/delete/list the dockerRegistry/kubernetes accounts.
3. Add the cloverctl to interate with the halyard daemon
Change-Id: I71bc5977f2d65aab88fa55f7d7a53ab75eb6a46b
Signed-off-by: wutianwei <wutianwei1@huawei.com>
|
|
|
|
|
|
|
|
- Uses client-go package to interface to k8s API and implement
functions as cloverkube package.
- Identifies GKE LB IP for clover-controller for user
- Identifies NodePort port number for clover-controller for user
if environment is local k8s (assumes flannel CNI currently)
- Deploys and deletes clover-collector and clover-controller with
native client-go constructs (currently images are defined with
local registry). Future work will implement other clover services
and Istio components. Uses the clover-system namespace.
- Uses Cobra go package to implement CLI (used in kubectl and
istioctl) using cloverctl <verb> <noun> convention.
- Interfaces to clover-controller to configure clover services
(visibility, IDS ...) within the cluster via REST messaging
- Start visibility (collector) engine using input yaml file or
defaults
- Init, stop and clear (truncate Cassandra tables) visibility
engine or get basic stats.
- Add custom rules to IDS from input yaml file and start/stop
IDS
- Generate jmeter testplan on jmeter-master using input yaml
file. Start tests and output log/results from CLI.
- Specify number of jmeter slaves to initiate tests on from
CLI. Automatically find IP addresses of jmeter slaves within
the k8s cluster.
- Sample yaml files for adding IDS rules, starting visibility
engine and generating jmeter test plans.
- Build script to install go and get dependent packages.
- Implement a custom Istio inject package for manual sidecar
injection (cloverinject). Currently, unused as it is built from
Istio 0.8.0/1.0.0 code base.
Change-Id: Ibb8d08cb98267bdffb8905c221473f177d51bbb3
Signed-off-by: Eddie Arrage <eddie.arrage@huawei.com>
|
|
|
|
|
|
- First pass of clover-controller which resides within the k8s
cluster and provides interfaces to all Clover services
- Only service that should need to be exposed outside of
cluster
- Docker build of container that uses stack of nginx, gunicorn
and flask to provide REST interface
- REST interface is intended to serve cloverctl CLI and
dashboard browser UI
- Implements GRPC messaging to clover-collector and snort
- GRPC interfaces files for snort/nginx are added to
container from repo. Collector GRPC files will be removed
from controller/control/api once patch below is merged
https://gerrit.opnfv.org/gerrit/#/c/57245/ and added
similarly
- Provides first pass callback for file upload from
clover-server.
- Some REST messages implement JSON for passing params
to internal services
- Redis interface added to obtain data from services.
Currently, a simple interface to retrieve snort event
information
- YAML manifest renderer to add to k8s. Uses NodePort
service currently, defaulting to port 32044.
- Removed collector gRPC interface files with merge of collector
- Expose tracing and monitoring host/port parameters, as these vary
depending on Istio version and Jaeger version
- Add logging to flask blueprints
- Added jmeter blueprint interface with REST for
testplan generation, start test and result retrieval
- Added flask Response to REST reply messages
- Retrieve some basic stats from collector in json
response
Change-Id: I59eaeb860445ade4b45bba22747a61fb0cf0bbd4
Signed-off-by: Eddie Arrage <eddie.arrage@huawei.com>
|
|
- Jmeter can be used for L4-7 functional and performance testing
- Jmeter master has gRPC server for management
- Generates Jmeter test plans from minimal yaml params file
(sample to be added with cloverctl) using template
- Optionally span tests across slave containers to allow greater
loads to be generated
- Specify loop/thread/slave count and URL list, which
dictates target and number of connections that will be attempted
- clover-controller will interface to gRPC interface on Jmeter
master
- Start tests on master and retrieve log/result files
- Render master and slave k8s manifests files
Change-Id: Id144c8f551b7d375ff252c8de0611f895b50387c
Signed-off-by: Eddie Arrage <eddie.arrage@huawei.com>
|
|
- Left the file samples/scenarios/service_delivery_controller_opnfv.yaml unchanged.
- Added a yaml definition of Cassandra StatefulSet and its service into a separate file under tools directory
- Cassandra Service run with 1 replica
- Deleted 'data-plane-ns' and use 'default' instead for cassandra containers.
- Revoked changes for samples/scenarios/service_delivery_controller_opnfv.yaml.
- Added new line (Wutien suggested it)
JIRA: CLOVER-000
Change-Id: I2bb4249cf2523f5011d6fefc69dc469a90e20eaf
Signed-off-by: iharijono <indra.harijono@huawei.com>
|
|
|
|
Change-Id: I0335fa912a3ca2dff5c989fa06183065216f10e4
Signed-off-by: wutianwei <wutianwei1@huawei.com>
|
|
if we set testid and start test immediately,
the first test's result can't be got from jaeger
Change-Id: Ia2ab8a91d8c5f9956ea4d3d7c2436fb05490acee
Signed-off-by: wutianwei <wutianwei1@huawei.com>
|
|
|
|
- Added a container named clover-collector using clover
container as a base with build script
- GRPC server to manage collector process
- Cassandra DB client interface to initialize visibility keyspace
- Init messaging adds table schemas for tracing - traces & spans
- Adds table for monitoring - metrics
- Does not implement Cassandra server but developed using
public Cassandra docker container
- Collector process in simple loop that periodically fetches
traces and monitoring data and inserts to Cassandra - not optimized
for batch retrieval yet for monitoring
- CLI interface added to collector process and used
by GRPC server for configuration
- Simple GRPC client script to test GRPC server and start/stop
of collector process
- Collector process can be configured with access for tracing,
monitoring and Cassandra
- Added a return value in monitoring query method
- Added ability to truncate tracing, metrics and spans tables
in cql
- Added cql prepared statements and batch insert for metrics
and spans
- Align cql connection to cql deployment within k8s
- Fix issue with cql host list using ast and collect process
args with background argument
- Added redis interface to accept service/metric list
externally for monitoring (will work in conjunction
with clover-controller)
- Use k8s DNS names and default ports for monitoring, tracing
and cassandra
- Added yaml manifest renderer/template for collector
Change-Id: I3e4353e28844c4ce9c185ff4638012b66c7fff67
Signed-off-by: Eddie Arrage <eddie.arrage@huawei.com>
|
|
Change-Id: Ib5b2240de3276164fe9e272bf36f0d1f89f409c0
Signed-off-by: Yujun Zhang <zhang.yujunz@zte.com.cn>
|
|
Change-Id: I6a1e526bec4160bcdac32d4124acb110b9cf6959
Signed-off-by: Yujun Zhang <zhang.yujunz@zte.com.cn>
|
|
the SDC application"
|
|
and on the SDC application
Change-Id: I6e1bd84a6d674a2c4c4484722b20415f5402a59c
Signed-off-by: Stephen Wong <stephen.kf.wong@gmail.com>
|
|
- cluster health is not red
- indics found
- log entry created by istio found
- requests in and out http load balance matches
pytest is used as the test runner and wrapped in `validate.py`
Change-Id: Iad540b69d05118fadc97df679cf3424513c15e38
Signed-off-by: Yujun Zhang <zhang.yujunz@zte.com.cn>
|
|
- Changed default Jaeger ports to 16686 for use with basic
kubernetes port-forward and CI scripts
- Added CLI to validate script to disable istio service check
by default. This requires at least a single http request
to istio-ingress after Jaeger deployment. It can be enabled
with 'python validate.py -s'. Port and IP address for Jaeger
can optionally be specified with '-ip' and '-port' options
- Modified tracing doc to add k8s port-forward example in addition
to k8s expose
Change-Id: I10fb4d3cccfa50370d44ed7446f67a49c538bba9
Signed-off-by: Eddie Arrage <eddie.arrage@huawei.com>
|
|
- Use a community yaml for redis in k8s as simple data store
- Redis can be used for tracing and also by the snort-ids
to store alerts that can be processed by other services
- If flannel is used, the redis CLI can be accessed on the
host OS with redis-cli -h <flannel ip>
- Within the k8s cluster, the redis service can be accessed with
DNS using name 'redis'
- The same yaml for redis is also included in toplevel manifest for SDC
scenario. Included here if intention is to use separately (tracing
only)
Change-Id: Ibad283a4cc8938fe01f5de6b7743bdb5511be3af
Signed-off-by: Eddie Arrage <eddie.arrage@huawei.com>
|
|
- install dependent deb/pip packages
- install basic tools istioctl, kubectl
- install clover source code
- build/upload docker image script
- update requirements.txt
- update module import path
- To use this image use need setup kube-config file.
e.g. `docker run -v /root/config:/root/.kube/config -it clover bash`
Change-Id: I91044bb99ce8e2b785ef03212d961a97b3d42233
Signed-off-by: QiLiang <liangqi1@huawei.com>
|
|
orchestration/kube_client, and tools/clover_validate_rr
Add an 'orchestration' directory. Please note that
'orchestration' does NOT mean Clover does any orchestration ---
similar to how Clover doesn't by itself implement tracing or
logging, orchestration is a directory for code related to Docker
orchestration client --- such as k8s client
kube_client utilizes the Kubernetes python client (a dependency)
to perform tasks against Kubernetes API server. For this commit,
it is only tested for weighted route rule verification, it does
three tasks:
(1) get a list of pods under a namespace --- pod dictionary now
only contains pod name and label dictionary: used to match
pod name with the node name in traces from OpenTracing
(2) check to see if a particular pod is up in a particular
namespace: used to check if Istio pods are running in
istio-system namespace
(3) check if a container exists in a list of pods under a
namespace: used to check if application pods have
istio-proxy container running
route_rule directly invokes istioctl as there isn't any Istio
Python client yet. Currently it reads and parses routerules
from Istio, and validates if a particular trace result matches
the routerules
Finally, a sample tool clover_validate_rr is provided. This
tool assumes a previous test has been ran (with an id with
both the route-rule-under-test and corresponding traces are
stored --- currently the assumption is tests were ran with
redis-master running on system). The tool can be invoked:
python clover_validate_rr.py -t <test-id> -s <service name>
where test-id is the ID of the test (most likely uuid) and
service name is the name of the service running in the
Kubernetes cluster upon which test traces should be fetched
against
Change-Id: Ic8ab6efc23c71ac4643bee796ef986a86f6fc7dd
Signed-off-by: Stephen Wong <stephen.kf.wong@gmail.com>
|
Stephen Wong
earrage
JingLu5
Cédric Ollivier
wutianwei
iharijono
Yujun Zhang
QiLiang