Artefatto

“ Forse l’amore é un artefatto di un’altra dimensione che non possiamo percepire consciamente… è l’unica cosa che possiamo percepire che trascende le dimensioni di tempo e spazio. Forse di questo dovremmo fidarci anche se non riusciamo a capirlo ancora. “

La La Land

Potrei ascoltare la stessa canzone per ore ed ore ed ogni volta farmi lo stesso viaggio!
Che bella la musica, che bello il pianoforte !

Lo splendore della disobbedienza

antigone

Prendi tua figlia, portala a Siracusa, siediti sui gradoni del teatro greco e insegnale lo splendore della disubbidienza. E’ rischioso ma è più rischioso non farlo mai.

Grabriele Romagnoli, dall’articolo “Cercasi Antigone per la Rivoluzione”, Vanity Fair del 5-06-2013, pag. 74

“Antigone”, disse sua sorella”Ti uccideranno”. “In qualche modo dovrò pur morire”, disse Antigone sempre rivolgendole le spalle. “Dovrò passare molto più tempo da morta che da viva”. “Sei pazza”, le disse la sorella.” E mi fai paura. Ma giuro. Non racconterò a nessuno quello che stai per fare”. L’affermazione fece girare la sorella minore per guardare di nuovo la maggiore negli occhi. ” URLALO PURE CON TUTTA LA VOCE CHE HAI!”, strillò.” NON MI IMPORTA CIO’ CHE DICI. NON MI IMPORTA CIO’ CHE CREONTE DICE. NON MI IMPORTA CIO’ CHE DICE LA GENTE, NON MI IMPORTA DI NESSUNO.”

Da “La storia di Antigone raccontata da Ali Smith” , 2011, Gruppo editoriale L’Espresso. Illustrazioni di Laura Paoletti.

Quando  mi chiedono perché ho chiamato mia figlia primogenita Antea non dico mai la verità perché temo di apparire troppo intellettuale radical chic, di fare la figura della snob (che non sono) e quindi rispondo delle ragioni che sono più o meno vere , ma che non sono LA ragione principale. Dico che l’ho scelto perché è un nome raro e originale ma al tempo stesso semplice e dolce, perché l’etimologia richiama la parola greca che significa “fiore” (anthos), perché il nonno paterno si chiama Antonio e i due nomi hanno la stessa radice, perché all’Università avevo una compagna di studi che ho poi perso di vista che si chiamava così ed era bellissima e intelligentissima. Ma la verità è che mia figlia si chiama così perché circa 2500 anni fa un tragediografo, Sofocle, ha reso immortale uno dei personaggi della mitologia greca più affascinanti: Antigone. Dicevano i latini che nel nome c’è il nostro destino (“nomen est omen”) ma non auguro di certo il destino di Antigone ad Antea: figlia di Edipo, che si cava gli occhi quando scopre di aver amato la madre, Antigone si oppone al divieto imposto dallo zio Creonte di seppellire il corpo del fratello Polinice che doveva giacere, privo di dignità e pietà umana, in pasto alle carogne e agli uccelli rapaci, perché si era opposto al fratello Eteocle, usurpatore del trono tebano. La ribellione di Antigone ad una legge iniqua, che va contro la legge di natura di dare degna sepoltura ai morti, le costerà la vita. Nel nome di mia figlia, così come in quello di Antigone, c’è questa parola: “anti”, questo seme di ribellione e di anticonformismo che spero germogli perché, come scrive Romagnoli nel suo articolo:

Tutte le istituzioni, per prima la famiglia, insegnano ai bambini il valore dell’obbedienza. Antigone insegna quello della disobbedienza. L’obbedienza porta un premio, la disobbedienza un castigo. Richiede coraggio, spirito di sacrificio, idealismo. Ma è la disobbedienza lo strappo che consente alla storia di avanzare sul solco della giustizia e non in quello della volontà di un qualunque governante.

Ecco perché mia figlia l’ho chiamata Antea: perché volevo che il nome le trasmettesse, per “osmosi”, proprio queste virtù che elenca Romagnoli: il coraggio, lo spirito di sacrificio, l’idealismo. Perché Antigone è Ipazia, che nel IV sec d. c. viene uccisa per difendere il suo amore per la scienza, è Malala Yousafzi,  che ha quasi perso la vita per affermare il diritto allo studio delle bambine pakistane, è Eleonora Pimental de Fonseca, è  Rosa Parks, è Dolores Ibàrruri, e tante altre ancora e, in questi giorni, è la ragazza vestita di rosso che al Gezi Park di Istanbul resta immobile di fronte al getto dell’idrante della polizia, simbolo dell’ accecante splendore della disobbedienza.

NGINX/PHP-FPM graceful shutdown and 502 errors

We have a PHP application running with Kubernetes in pods with two dedicated containers — NGINX и PHP-FPM.

The problem is that during downscaling clients get 502 errors. E.g. when a pod is stopping, its containers can not correctly close existing connections.

So, in this post, we will take a closer look at the pods’ termination process in general, and NGINX and PHP-FPM containers in particular.

Testing will be performed on the AWS Elastic Kubernetes Service by the Yandex.Tank utility.

Ingress resource will create an AWS Application Load Balancer with the AWS ALB Ingress Controller.

Для управления контейнерами на Kubernetes WorkerNodes испольузется Docker.

Pod Lifecycle — Termination of Pods

So, let’s take an overview of the pods’ stopping and termination process.

Basically, a pod is a set of processes running on a Kubernetes WorkerNode, which are stopped by standard IPC (Inter-Process Communicationsignals.

To give the pod the ability to finish all its operations, a container runtime at first ties softly stop it (graceful shutdown) by sending a SIGTERM signal to a PID 1 in each container of this pod (see docker stop). Also, a cluster starts counting a grace period before force kill this pod by sending a SIGKILL signal.

The SIGTERM can be overridden by using the STOPSIGNAL in an image used to spin up a container.

Thus, the whole flow of the pod’s deleting is (actually, the part below is a kinda copy of the official documentation):

  1. a user issues a kubectl delete pod or kubectl scale deployment command which triggers the flow and the cluster start countdown of the grace period with the default value set to the 30 second
  2. the API server of the cluster updates the pod’s status — from the Running state, it becomes the Terminating (see Container states). A kubelet on the WorkerNode where this pod is running, receives this status update and start the pod’s termination process:
  3. if a container(s) in the pod have a preStop hook – kubelet will run it. If the hook is still running the default 30 seconds on the grace period – another 2 seconds will be added. The grace period can be set with the terminationGracePeriodSeconds
  4. when a preStop hook is finished, a kubelet will send a notification to the Docker runtime to stop containers related to the pod. The Docker daemon will send the SIGTERM signal to a process with the PID 1 in each container. Containers will get the signal in random order.
  5. simultaneously with the beginning of the graceful shutdown — Kubernetes Control Plane (its kube-controller-manager) will remove the pod from the endpoints (see Kubernetes – Endpoints) and a corresponding Service will stop sending traffic to this pod
  6. after the grace period countdown is finished, a kubelet will start force shutdown – Docker will send the SIGKILL signal to all remaining processes in all containers of the pod which can not be ignored and those process will be immediately terminated without change to correctly finish their operations
  7. kubelet triggers deletion of the pod from the API server
  8. API server deletes a record about this pod from the etcd
Image for post

Actually, there are two issues:

  1. the NGINX and PHP-FPM perceives the SIGTERM signal as a force как “brutal murder” and will finish their processes immediately , и завершают работу немедленно, without concern about existing connections (see Controlling nginx and php-fpm(8) – Linux man page)
  2. the 2 and 3 steps — sending the SIGTERM and an endpoint deletion – are performed at the same time. Still, an Ingress Service will update its data about endpoints not momentarily and a pod can be killed before then an Ingress will stop sending traffic to it causing 502 error for clients as the pod can not accept new connections

E.g. if we have a connection to an NGINX server, the NGINX master process during the fast shutdown will just drop this connection and our client will receive the 502 error, see the Avoiding dropped connections in nginx containers with “STOPSIGNAL SIGQUIT”.

NGINX STOPSIGNAL and 502

Okay, now we got some understanding of how it’s going — let’s try to reproduce the first issue with NGINX.

The example below is taken from the post above and will be deployed to a Kubernetes cluster.

Prepare a Dockerfile:

FROM nginx

RUN echo 'server {\n\
listen 80 default_server;\n\
location / {\n\
proxy_pass http://httpbin.org/delay/10;\n\
}\n\
}' > /etc/nginx/conf.d/default.conf

CMD ["nginx", "-g", "daemon off;"]

Here NGINX will proxy_pass a request to the http://httpbin.org which will respond with a 10 seconds delay to emulate a PHP-backend.

Build an image and push it to a repository:

$ docker build -t setevoy/nginx-sigterm .
$ docker push setevoy/nginx-sigterm

Now, add a Deployment manifest to spin up 10 pods from this image.

Here is the full file with a Namespace, Service, and Ingress, in the following part of this post, will add only updated parts of the manifest:

---
apiVersion: v1
kind: Namespace
metadata:
name: test-namespace
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: test-deployment
namespace: test-namespace
labels:
app: test
spec:
replicas: 10
selector:
matchLabels:
app: test
template:
metadata:
labels:
app: test
spec:
containers:
- name: web
image: setevoy/nginx-sigterm
ports:
- containerPort: 80
resources:
requests:
cpu: 100m
memory: 100Mi
readinessProbe:
tcpSocket:
port: 80
---
apiVersion: v1
kind: Service
metadata:
name: test-svc
namespace: test-namespace
spec:
type: NodePort
selector:
app: test
ports:
- protocol: TCP
port: 80
targetPort: 80
---
apiVersion: extensions/v1beta1
kind: Ingress
metadata:
name: test-ingress
namespace: test-namespace
annotations:
kubernetes.io/ingress.class: alb
alb.ingress.kubernetes.io/scheme: internet-facing
alb.ingress.kubernetes.io/listen-ports: '[{"HTTP": 80}]'
spec:
rules:
- http:
paths:
- backend:
serviceName: test-svc
servicePort: 80

Deploy it:

$ kubectl apply -f test-deployment.yaml
namespace/test-namespace created
deployment.apps/test-deployment created
service/test-svc created
ingress.extensions/test-ingress created

Check the Ingress:

$ curl -I aadca942-testnamespace-tes-5874–698012771.us-east-2.elb.amazonaws.com
HTTP/1.1 200 OK

And we have 10 pods running:

$ kubectl -n test-namespace get pod
NAME READY STATUS RESTARTS AGE
test-deployment-ccb7ff8b6–2d6gn 1/1 Running 0 26s
test-deployment-ccb7ff8b6–4scxc 1/1 Running 0 35s
test-deployment-ccb7ff8b6–8b2cj 1/1 Running 0 35s
test-deployment-ccb7ff8b6-bvzgz 1/1 Running 0 35s
test-deployment-ccb7ff8b6-db6jj 1/1 Running 0 35s
test-deployment-ccb7ff8b6-h9zsm 1/1 Running 0 20s
test-deployment-ccb7ff8b6-n5rhz 1/1 Running 0 23s
test-deployment-ccb7ff8b6-smpjd 1/1 Running 0 23s
test-deployment-ccb7ff8b6-x5dc2 1/1 Running 0 35s
test-deployment-ccb7ff8b6-zlqxs 1/1 Running 0 25s

Prepare a load.yaml for the Yandex.Tank:

phantom:
address: aadca942-testnamespace-tes-5874-698012771.us-east-2.elb.amazonaws.com
header_http: "1.1"
headers:
- "[Host: aadca942-testnamespace-tes-5874-698012771.us-east-2.elb.amazonaws.com]"
uris:
- /
load_profile:
load_type: rps
schedule: const(100,30m)
ssl: false
console:
enabled: true
telegraf:
enabled: false
package: yandextank.plugins.Telegraf
config: monitoring.xml

Here, we will perform 1 request per second to pods behind our Ingress.

Run tests:

Image for post

All good so far.

Now, scale down the Deployment to only one pod:

$ kubectl -n test-namespace scale deploy test-deployment — replicas=1
deployment.apps/test-deployment scaled

Pods became Terminating:

$ kubectl -n test-namespace get pod
NAME READY STATUS RESTARTS AGE
test-deployment-647ddf455–67gv8 1/1 Terminating 0 4m15s
test-deployment-647ddf455–6wmcq 1/1 Terminating 0 4m15s
test-deployment-647ddf455-cjvj6 1/1 Terminating 0 4m15s
test-deployment-647ddf455-dh7pc 1/1 Terminating 0 4m15s
test-deployment-647ddf455-dvh7g 1/1 Terminating 0 4m15s
test-deployment-647ddf455-gpwc6 1/1 Terminating 0 4m15s
test-deployment-647ddf455-nbgkn 1/1 Terminating 0 4m15s
test-deployment-647ddf455-tm27p 1/1 Running 0 26m

And we got our 502 errors:

Image for post

Next, update the Dockerfile — add the STOPSIGNAL SIGQUIT:

FROM nginx

RUN echo 'server {\n\
listen 80 default_server;\n\
location / {\n\
proxy_pass http://httpbin.org/delay/10;\n\
}\n\
}' > /etc/nginx/conf.d/default.conf

STOPSIGNAL SIGQUIT

CMD ["nginx", "-g", "daemon off;"]

Build, push:

$ docker build -t setevoy/nginx-sigquit .
docker push setevoy/nginx-sigquit

Update the Deployment with the new image:

...
spec:
containers:
- name: web
image: setevoy/nginx-sigquit
ports:
- containerPort: 80
...

Redeploy, and check again.

Run tests:

Image for post

Scale down the deployment again:

$ kubectl -n test-namespace scale deploy test-deployment — replicas=1
deployment.apps/test-deployment scaled

And no errors this time:

Image for post

Great!

Traffic, preStop, and sleep

But still, if repeat tests few times we still can get some 502 errors:

Image for post

This time most likely we are facing the second issue — endpoints update is performed at the same time when the SIGTERM Is sent.

Let’s add a preStop hook with the sleep to give some time to update endpoints and our Ingress, so after the cluster will receive a request to stop a pod, a kubelet on a WorkerNode will wait for 5 seconds before sending the SIGTERM:

...
spec:
containers:
- name: web
image: setevoy/nginx-sigquit
ports:
- containerPort: 80
lifecycle:
preStop:
exec:
command: ["/bin/sleep","5"]
...

Repeat tests — and now everything is fine

Our PHP-FPM had no such issue as its image was initially built with the STOPSIGNAL SIGQUIT.

Other possible solutions

And of course, during debugging I’ve tried some other approaches to mitigate the issue.

See links at the end of this post and here I’ll describe them in short terms.

preStop and nginx -s quit

One of the solutions was to add a preStop hook which will send QUIT to NGINX:

lifecycle:
preStop:
exec:
command:
- /usr/sbin/nginx
- -s
- quit

Or:

...
lifecycle:
preStop:
exec:
command:
- /bin/sh
- -SIGQUIT
- 1
....

But it didn’t help. Not sure why as the idea seems to be correct — instead of waiting for the TERM from Kubernetes/Docker – we gracefully stopping the NGINX master process by sending QUIT.

You can also run the strace utility check which signal is really received by the NGINX.

NGINX + PHP-FPM, supervisord, and stopsignal

Our application is running in two containers in one pod, but during the debugging, I’ve also tried to use a single container with both NGINX and PHP-FPM, for example, trafex/alpine-nginx-php7.

There I’ve tried to add to stopsignal to the supervisor.conf for both NGINX and PHP-FPM with the QUIT value, but this also didn’t help although the idea also seems to be correct.

Still, one can try this way.

PHP-FPM, and process_control_timeout

In the Graceful shutdown in Kubernetes is not always trivial and on the Stackoveflow in the Nginx / PHP FPM graceful stop (SIGQUIT): not so graceful question is a note that FPM’s master process is killed before its child and this can lead to the 502 as well.

Not our current case, but pay your attention to the process_control_timeout.

NGINX, HTTP, and keep-alive session

Also, it can be a good idea to use the [Connection: close] header – then the client will close its connection right after a request is finished and this can decrease 502 errors count.

But anyway they will persist if NGINX will get the SIGTERM during processing a request.

See the HTTP persistent connection.

Deploying Kubernetes on bare metal with Rancher 2.0

Contents

  • Install Rancher server
  • Create a Kubernetes cluster
  • Add Kubernetes nodes
  • Install StorageOS as the Kubernetes storage class
  • Understand Nginx Ingress in Rancher

Install Rancher

Create a VM with Docker and Docker Compose installed and install Rancher 2.0 with docker compose:

  • Rancher docker-compose file: docker-compose.yaml
  • Run these commands to install Rancher with docker compose:
    • git clone https://github.com/polinchw/rancher-docker-compose
    • cd rancher-docker-compose
    • docker-compose up -d

Create your Kubernetes cluster with Rancher

Install a custom Kubernetes cluster with Rancher. Use the ‘Custom’ cluster.

Cluster!

Add Kubernetes nodes and join the Kubernetes cluster

Run the following commands on all the VMs that your Kubernetes cluster will run on. The final docker command will have the VM join the new Kubernetes cluster.

Replace the –server and –token with your Rancher server and cluster token.

#!/bin/bash

#sudo apt update
#sudo apt -y dist-upgrade

#Ubuntu (Docker install)
#sudo apt -y install docker.io

sudo apt -y install linux-image-extra-$(uname -r)

#Debian 9 (Docker install)
#sudo apt -y install apt-transport-https ca-certificates curl gnupg2 software-properties-common
#curl -fsSL https://download.docker.com/linux/debian/gpg | sudo apt-key add -
#sudo add-apt-repository "deb [arch=amd64] https://download.docker.com/linux/debian $(lsb_release -cs) stable"
#sudo apt update
#sudo apt -y install docker-ce

sudo mkdir -p /etc/systemd/system/docker.service.d/
sudo cat <<EOF > /etc/systemd/system/docker.service.d/mount_propagation_flags.conf
[Service]
MountFlags=shared
EOF

sudo systemctl daemon-reload
sudo systemctl restart docker.service

#This is dependent on your Rancher server
sudo docker run -d --privileged --restart=unless-stopped --net=host -v /etc/kubernetes:/etc/kubernetes -v /var/run:/var/run rancher/rancher-agent:v2.1.0-rc9 --server https://75.77.159.159 --token rb8k8kkqw55jqnqbbf4ssdjqtw6hndhfxxcghgv8257kx4p6qsqq55 --ca-checksum 641b2888ce3f1091d20149a495d10457154428f440475b42291b6af1b6c0dd06 --etcd --controlplane --worker

Download the kub config file for the cluster

Helloservice!

After you download the kub config file you can use it by running this command:

export KUBECONFIG=$HOME/.kube/rancher-config

Install Helm on the cluster

git clone https://github.com/polinchw/set-up-tiller

cd set-up-tiller

chmod u+x set-up-tiller.sh

./set-up-tiller.sh

helm init --service-account tiller


Install StorageOS Helm Chart

helm repo add storageos https://charts.storageos.com
helm install --name storageos --namespace storageos-operator --version 1.1.3 storageos/storageoscluster-operator

Add the Storage OS Secret

apiVersion: v1
kind: Secret
metadata:
  name: storageos-api
  namespace: default
  labels:
    app: storageos
type: kubernetes.io/storageos
data:
  # echo -n '<secret>' | base64
  apiUsername: c3RvcmFnZW9z
  apiPassword: c3RvcmFnZW9z


Add the StorageOSCluster

apiVersion: storageos.com/v1
kind: StorageOSCluster
metadata:
  name: example-storageos
  namespace: default
spec:
  secretRefName: storageos-api
  secretRefNamespace: default
  csi:
    enable: true


Set StorageOS as the default storage class

kubectl patch storageclass fast -p ‘{“metadata”: {“annotations”:{“storageclass.kubernetes.io/is-default-class”:”true”}}}’

Using the default Nginx Igress

Rancher automatically installs the nginx ingress controller on all the nodes in the cluster.
If you are able to expose one of the VMs in the cluster to the outside world with a public IP then you can connect to the ingress based services on ports 80 and 443.

Any app you want to be accessible through the default nginx ingress must be added to the ‘default’ project in Rancher.