featurebase/dax/computer/service/computer.go
Travis Turner a44b622aa0 Introduce ServiceManager and Refactor DAX Integration tests (#2320)
* Introduce ServiceManager and Refactor DAX Integration tests

The ServiceManager provides an interface with which to manage
featurebase (dax) services (mds, queryer, computer). It replaces the
confusing interface implementations in /dax/server/server.go (which
optionally used pointers to in-process objects to satisfy an interface)
with (for now) http implementations. The thought is that even if we're
running all services in-process, we should communicate between services
over http in order to mirror what we would do in a production
environment where the services are running on different nodes.

This batch of commits does quit a lot, most of which is captured here:

- Added `path` support to `dax.Address`. Address is now a string of the form [scheme]://[host]:[port]/[path].
- Added `Holder.directiveApplied` to determine (in tests) if the computer has completed applying the latest directive. This is somewhat temporary until we improve the mds-to-computer logic.
- Removed the "service prefix" code which was prepending client URL paths with the prefix. Instead, the serviceType (mds, queryer, computer[n] is now part of `dax.Address`).
- Removed, from the dax config, the top level `StorageMethod` and `StorageDSN` and now just have `MDS.Config.DataDir`.
- Added `Computer.Config.N` to specify the number of computers to run in-process.
- Moved the `pilosa.MDS` interface to `computer.Registrar`. This is an example of getting the interfaces defined in the right packages.
- Added `SnapshotTable()` method to the mds client (to align with its API).
- Changed `Balancer.AddJob()` to `Balancer.AddJobs()` to support, for example, adding 256 partitions in a single call. Refactored some of the naive Balancer to account for this.
- Added a `Seed` to the top-level config. It's not really useful because of package `crypto/rand`.
- Added an in-memory implementation of the DisCo interface and disabled etcd in a computer service.
- Create sepearte data-dirs for each in-process computer.
- Disabled grpc in dax.
- Modified the sql3 test definition format to support multiple insert steps and separate query results (to align with those steps).

* Changes necessary to get multiple computer instance running in-process

For now the config looks like this:

```
[computer]
run = true
n = 4
```

but we can probably just change that to be something like:

```
[computer]
run = 4
```

*Issues found running multiple "computers" in-process*
- grpc was trying to bind on the same port
  - changed GRPCListener from `*net.TCPListener` to `net.Listener`
  - created a nopListener and set to that for now (i.e. disabled grpc)
- etcd was starting more than once
  - changed dax to use in-memory implementations of the disco interfaces (i.e. stop using etcd)
- IDAllocator (which uses boltdb) was trying to open the `idalloc.db` file more than once
  - realized we have to set separate data-dirs for each holder. that fixed it.

* Port dax integration tests to ManagedCommand

* Modify Balancer-related methods like AddJob to AddJobs

There were (and still are) a lot of places where we were adding on job
at a time, even when we had a long list of jobs to add. This resulted in
every job add (for example adding 1 of 256 shards) taking ~40ms, or over
10s to create a keyed table. One reason was because each job add was
making multiple boltdb transactions.

* Port over more dax integration test stuff

* Add DirectiveApplied to signify that snapshot/writes have loaded.

We use this in tests to avoid using sleeps.
This should be considered temporary; we're going to need a more robust
solution for determining when a computer node is ready to serve complete
data.

* Finish porting dax integration tests

* Improve godocs

* Remove docker-based DAX integration tests.

* go mod tidy

* Move test/managed.go to avoid package conflicts

* Modify IDK integration tests to work with ServiceManager changes

This is really just computer -> computer0
And the MDS DataDir config change.

* cleanup found during review

* echo $CI_COMMIT_REF_SLUG in CI

* remove docker image arg, use build instead

(cherry picked from commit 2843f218bc)
2022-12-12 09:01:20 -08:00

196 lines
5.7 KiB
Go

package service
import (
"context"
"io"
"net"
"net/http"
featurebase "github.com/molecula/featurebase/v3"
"github.com/molecula/featurebase/v3/dax"
"github.com/molecula/featurebase/v3/dax/computer"
mdsclient "github.com/molecula/featurebase/v3/dax/mds/client"
"github.com/molecula/featurebase/v3/dax/snapshotter"
snapshotterclient "github.com/molecula/featurebase/v3/dax/snapshotter/client"
"github.com/molecula/featurebase/v3/dax/writelogger"
writeloggerclient "github.com/molecula/featurebase/v3/dax/writelogger/client"
"github.com/molecula/featurebase/v3/errors"
"github.com/molecula/featurebase/v3/logger"
fbserver "github.com/molecula/featurebase/v3/server"
)
// Ensure type implements interface.
var _ dax.ComputerService = (*computerService)(nil)
type computerService struct {
addr dax.Address
cfg CommandConfig
key dax.ServiceKey
computer *fbserver.Command
logger logger.Logger
}
func New(addr dax.Address, cfg CommandConfig, logger logger.Logger) *computerService {
cfg.ComputerConfig.Advertise = addr.HostPort()
return &computerService{
addr: addr,
cfg: cfg,
logger: logger,
}
}
func (c *computerService) Start() error {
// We initialize the fbserver.Command at start (as opposed to in New()
// above) because we want to inject (via cfg.Name) the ServiceKey into the
// *fbserver.Command returned by newCommand() so that it can be used
// internally. In the future, when we have the pilosa package moved down to
// the computer package, we should be able to clean this up so that things
// happen in a reasonable order like we do with the other service types.
if c.computer == nil {
c.cfg.Name = string(c.Key())
c.computer = newCommand(c.addr, c.cfg)
if c.cfg.ComputerConfig.MDSAddress != "" {
mdsAddr := dax.Address(c.cfg.ComputerConfig.MDSAddress)
// Set mds (registrar) on computer.
if err := c.SetMDS(mdsAddr); err != nil {
return errors.Wrapf(err, "setting mds service on computer: %s, %v", c.cfg.Name, err)
}
}
}
if err := c.computer.StartNoServe(c.Address()); err != nil {
return errors.Wrap(err, "starting featurebase command")
}
if c.computer.Registrar != nil {
node := &dax.Node{
Address: c.Address(),
RoleTypes: []dax.RoleType{
dax.RoleTypeCompute,
dax.RoleTypeTranslate,
},
}
if err := c.computer.Registrar.RegisterNode(context.TODO(), node); err != nil {
return errors.Wrapf(err, "registering computer: %s", c.Address())
}
}
return nil
}
func (c *computerService) Stop() error {
return c.computer.Close()
}
func (c *computerService) Key() dax.ServiceKey {
return c.key
}
func (c *computerService) SetKey(key dax.ServiceKey) {
c.key = key
}
func (c *computerService) Address() dax.Address {
return dax.Address(c.addr.HostPort() + "/" + string(c.key))
}
func (c *computerService) HTTPHandler() http.Handler {
return c.computer.HTTPHandler()
}
func (c *computerService) SetMDS(addr dax.Address) error {
c.computer.Registrar = mdsclient.New(addr, c.logger)
return nil
}
type CommandConfig struct {
// Name is used to distinguish between locally running commands.
// For example, it's appended to DataDir so that each cmd has a
// separate data directory for its holder.
Name string
WriteLoggerRun bool
WriteLoggerConfig writelogger.Config
SnapshotterRun bool
SnapshotterConfig snapshotter.Config
ComputerConfig fbserver.Config
Listener net.Listener
RootDataDir string
Stderr io.Writer
Logger logger.Logger
}
func newCommand(addr dax.Address, cfg CommandConfig) *fbserver.Command {
// Set up WriteLogger.
// TODO(tlt): since WriteLogger is no longer a separate service (but
// rather just a directory path) its configuration could be moved under
// computer, and then get rid of WriteLogger.Run. This would become "if
// DataDir != ''". Let's do this after we get rid of the dax integration
// tests which start up separate writelogger and snapshotter containers.
var wlSvc *writelogger.WriteLogger
if cfg.WriteLoggerRun {
wlSvc = writelogger.New(writelogger.Config{
DataDir: cfg.WriteLoggerConfig.DataDir,
Logger: cfg.Logger,
})
}
// Set up Snapshotter.
var ssSvc *snapshotter.Snapshotter
if cfg.SnapshotterRun {
ssSvc = snapshotter.New(snapshotter.Config{
DataDir: cfg.SnapshotterConfig.DataDir,
Logger: cfg.Logger,
})
}
// Set the FeatureBase.Config values based on the top-level Config
// values.
cfg.ComputerConfig.Listener = &nopListener{}
cfg.ComputerConfig.Advertise = addr.HostPort()
cfg.ComputerConfig.GRPCListener = &nopListener{}
cfg.ComputerConfig.DataDir = cfg.RootDataDir + "/" + cfg.Name
var writeLoggerImpl computer.WriteLogger
if cfg.ComputerConfig.WriteLogger != "" {
writeLoggerImpl = writeloggerclient.New(dax.Address(cfg.ComputerConfig.WriteLogger))
} else if wlSvc != nil {
writeLoggerImpl = wlSvc
} else {
cfg.Logger.Warnf("No writelogger configured, dynamic scaling will not function properly.")
}
var snapshotterImpl computer.Snapshotter
if cfg.ComputerConfig.Snapshotter != "" {
snapshotterImpl = snapshotterclient.New(dax.Address(cfg.ComputerConfig.Snapshotter))
} else if ssSvc != nil {
snapshotterImpl = ssSvc
} else {
cfg.Logger.Warnf("No snapshotter configured.")
}
fbcmd := fbserver.NewCommand(cfg.Stderr,
fbserver.OptCommandSetConfig(&cfg.ComputerConfig),
fbserver.OptCommandServerOptions(
featurebase.OptServerIsComputeNode(true),
featurebase.OptServerLogger(cfg.Logger),
),
fbserver.OptCommandInjections(fbserver.Injections{
WriteLogger: writeLoggerImpl,
Snapshotter: snapshotterImpl,
IsComputeNode: true,
}),
)
return fbcmd
}
type nopListener struct{}
func (n *nopListener) Accept() (net.Conn, error) { return nil, nil }
func (n *nopListener) Close() error { return nil }
func (n *nopListener) Addr() net.Addr { return nil }