multiple instances for one service; deployment scripts generator

This PR is related to issues https://gitlab.mpi-sws.org/cld/tracing/hindsight-grpc/-/issues/7 and https://gitlab.mpi-sws.org/cld/tracing/hindsight-grpc/-/issues/15.

config/gen_configs.py

+import math
+import json
+
+FACTOR = 0.6
+ROOT_FACTOR = 2
+
+# todo: separate the server list into a config file
+SERVERS = {'brain18': 48, 'brain19': 48, 'brain20': 48, 'pinky10': 16, 'pinky11': 16, 'pinky12': 16, 'pinky13': 16,
+           'pinky14': 16, 'pinky15': 16, 'pinky16': 16, 'pinky17': 16, 'pinky18': 16, 'pinky19': 16}
+
+
+TMUX = """
+send() {
+    tmux send "$1" ENTER;
+}
+
+server_setup () {
+    tmux select-pane -t $1
+    send "ssh $2"
+    send "docker load -i ~/docker_images/tracer_image.tar"
+}
+
+server_hindsight () {
+    send "docker run -it -v /home/zxie/hindsight-grpc/config/alibaba:/config/ \\
+        -d --name $1 --shm-size=2gb -p $2:$2 -p $5:$5 --rm tracer bash -c \\
+        \\"cd build; ./server -x $3 -c 128 -a /config/sub_alibaba_addresses.json \\
+        -t /config/sub_alibaba_topology.json -i $4 $1 \\
+        & sleep 10; cd /root/hindsight/agent/; \\
+        go run cmd/agent2/main.go -serv $1 -port $5 -lc pinky09:5252 -r pinky09:5253\\""
+}
+
+server_run () {
+    tmux select-pane -t $1
+    if [ $4 = "hindsight" ] || [ $4 = "ot-hindsight" ]
+    then
+        server_hindsight $2 $3 $4 $5 $6
+    else
+        send "docker run -it -v /home/zxie/hindsight-grpc/config/alibaba:/config/ \\
+            -d --name $2 -p $3:$3 --rm tracer bash -c \\
+            \\"cd build; ./server -x $4 -c 128 -a /config/sub_alibaba_addresses.json \\
+            -t /config/sub_alibaba_topology.json -h pinky09 -i $5 -p $7 $2\\""
+    fi
+}
+
+setup_servers () {
+__setup__
+}
+
+deploy_servers () {
+__deploy__
+}
+
+kill_servers () {
+__kill__
+}
+"""
+
+# directly change the policy in the generated config file
+OTEK_CONFIG = """
+receivers:
+__jaegers__
+
+processors:
+  tail_sampling:
+    decision_wait: 10s
+    num_traces: 10000000
+    policies: [
+        # current implementation is OR combination
+        # the trace will get sampled if any policy is triggered
+        { name: always_sample, type: always_sample },
+        # { name: latency, type: latency, latency: { threshold_ms: 100 } },
+        # {
+        #   name: probabilistic,
+        #   type: probabilistic,
+        #   probabilistic: { sampling_percentage: 0.1 },
+        # },
+        # {
+        #   name: status_code,
+        #   type: status_code,
+        #   status_code: { status_codes: [ERROR] },
+        # },
+        # {
+        #   # this one is not well defined
+        #   name: rate_limiting,
+        #   type: rate_limiting,
+        #   rate_limiting: { spans_per_second: 35 },
+        # },
+    ]
+
+exporters:
+  file:
+    path: ./gateway.json
+
+  logging:
+    loglevel: info
+    # number of messages initially logged each second.
+    sampling_initial: 5
+    # sampling rate after the initial messages are logged (every Mth message is logged)
+    sampling_thereafter: 5
+
+service:
+  pipelines:
+    traces:
+      receivers: [__receivers__]
+      processors: [tail_sampling]
+      exporters: [file, logging]
+"""
+
+jaeger = \
+    """  jaeger/{}:
+    protocols:
+      thrift_compact:
+        endpoint: 0.0.0.0:{}
+"""
+
+
+def mapping_services(service_pair, root):
+    server_plan = {s: {} for s in SERVERS}
+
+    cores_all = sum(SERVERS.values())
+    service_all = sum([i[1] for i in service_pair])
+
+    for service, estimation in service_pair[::-1]:
+        server_pairs = sorted(SERVERS.items(), key=lambda item: item[1])
+        servers = [i[0] for i in server_pairs]
+        quota = [i[1] for i in server_pairs]
+
+        needed_cores = math.ceil(estimation / service_all * cores_all * FACTOR)
+        if service == root:
+            # extra resources for root service
+            needed_cores *= ROOT_FACTOR
+        # needed_cores = round(estimation / service_all * cores_all * FACTOR)
+        assert sum(quota) >= needed_cores and "resource shortage"
+
+        # no more than 5 instance per service
+        for dup in range(1, 6):
+            if min(quota[-dup:]) >= int(needed_cores/dup):
+                for i, s in enumerate(servers[-dup:]):
+                    server_plan[s][service] = int(needed_cores / dup)
+                    SERVERS[s] -= server_plan[s][service]
+                break
+        else:
+            assert False and "failed to allocate sufficient resources"
+
+    print("{:.2f}% of cores are allocated".format(
+        1 - sum(SERVERS.values()) / cores_all))
+    return server_plan
+
+
+def write_configs(server_plan):
+    sub_addr = []
+    pointer = 0
+    otel_port = 6832
+
+    setup_servers = ""
+    deploy_servers = ""
+    kill_servers = ""
+    jaegers = ""
+    receivers = []
+
+    service_entry = {}
+
+    for server in server_plan:
+        port = 50050
+        agent_port = 5050
+        for service in server_plan[server]:
+            instance_id = 0
+            if service in service_entry:
+                instance_id = len(
+                    sub_addr[service_entry[service]]["instances"])
+                sub_addr[service_entry[service]]["instances"].append(
+                    {"hostname": server, "port": str(port), "agent_port": str(agent_port)})
+            else:
+                sub_addr.append({"name": service, "deploy_addr": "0.0.0.0", "instances": [
+                                {"hostname": server, "port": str(port), "agent_port": str(agent_port)}]})
+                service_entry[service] = len(service_entry)
+            deploy_servers += "\tserver_run {} {} {} {} {} {} {}\n".format(
+                pointer, service, port, "$1", instance_id, agent_port, otel_port)
+            jaegers += jaeger.format(otel_port, otel_port)
+            receivers.append("jaeger/"+str(otel_port))
+            port += 1
+            agent_port += 1
+            otel_port += 1
+        setup_servers += "\tserver_setup {} {}\n".format(pointer, server)
+        kill_servers += "\ttmux select-pane -t {};\tsend \"docker stop \$(docker ps -a -q)\"\n".format(
+            pointer)
+        pointer += 1
+
+    tmux = TMUX.replace("__setup__", setup_servers)
+    tmux = tmux.replace("__deploy__", deploy_servers)
+    tmux = tmux.replace("__kill__", kill_servers)
+    otel_config = OTEK_CONFIG.replace("__jaegers__", jaegers)
+    otel_config = otel_config.replace("__receivers__", ",".join(receivers))
+
+    with open("alibaba/sub_alibaba_addresses.json", "w") as f:
+        json.dump({"addresses": sub_addr}, f, indent=2)
+
+    with open("alibaba/tmux_launch.sh", "w") as f:
+        f.write(tmux)
+
+    with open("alibaba/sub_alibaba_otel_config.yaml", "w") as f:
+        f.write(otel_config)