Repeatedly collect stats from the specified docker container until the stopSignal is closed or the context is cancelled

createImageCache scans the driver config to spawn an image cacher if applicable

TODO: I will remove this and other debug commented lines once I'm sure it's all right.

currently unused. But at some point, we need to age/delete old slotQueues.

uses grpc connection backoff protocol https://github.com/grpc/grpc/blob/master/doc/connection-backoff.md

removes docker err formatting: 'API Error (code) {"message":"..."}'

finalize initiates a graceful shutdown of the session. This is currently achieved by a sentinel nil enqueue to gRPC sender.

getDataMsg fetches/waits for a DataFrame message from the LB using inQueue (gRPC receiver)

headroom estimation in order not to consume entire RAM if possible

determine sidecar-monitored cmd names. But by default we track dockerd + containerd

TODO plumbing considerations, we've put the S pipe next to the chandalier...

getSlot returns a Slot (or error) for the request to run. This will wait for other containers to become idle or it may wait for resources to become av

getSlot must ensure that if it receives a slot, it will be returned, otherwise a container will be locked up forever waiting for slot to free.

getTryMsg fetches/waits for a TryCall message from the LB using inQueue (gRPC receiver)

handleTriggerHTTPFunctionCall2 executes the function and returns an error Requires the following in the context:

Handles a status call concurrency and caching.

handleTriggerHTTPFunctionCall2 executes the function and returns an error Requires the following in the context:

handleTryCall based on the TryCall message, tries to place the call on NBIO Agent

hotLauncher is spawned in a go routine for each slot queue to monitor stats and launch hot containers if needed. Upon shutdown or activity timeout, ho

initDockerMeasures initializes Docker related measures

Installs a child process reaper if init process

isGRPCError inspect the error to verify if it is gRPC status and return a bool as results of this operation true means the error is a gRPC error false

isSyslogError checks if the error message is what docker syslog plugin returns when not able to connect to syslog

isTooBusy checks if the error is a retriable error (503) that is explicitly sent by runner. If isTooBusy returns true then we can idempotently run thi

A Fast, Minimal Memory, Consistent Hash Algorithm: https://arxiv.org/ftp/arxiv/papers/1406/1406.2294.pdf

killLeakedContainers scans and destroys previously left over containers that were managed by this docker driver. This operation is executed once and i

latest version will find the latest version from a list of migration names (not from the db)

listenEventLoop listens for docker events and reconnects if necessary

listenEvents registers an event listener to docker to stream docker events and records these in stats.

TODO: switch to github.com/docker/engine-api