7121: Log all errors (except the everyday "not found") encountered during CompareAndT...

[arvados.git] / services / keepstore / keepstore.go

diff --git a/services/keepstore/keepstore.go b/services/keepstore/keepstore.go
index c6cb00db0fa28b621dc306e9a1d7ec7775c41282..53cf7be6ca09c62828787f1c9229cce13193b480 100644 (file)
--- a/services/keepstore/keepstore.go
+++ b/services/keepstore/keepstore.go
@@ -14,6 +14,7 @@ import (
        "os"
        "os/signal"
        "strings"
+       "sync"
        "syscall"
        "time"
 )
@@ -54,7 +55,10 @@ var data_manager_token string
 
 // never_delete can be used to prevent the DELETE handler from
 // actually deleting anything.
-var never_delete = false
+var never_delete = true
+
+var maxBuffers = 128
+var bufs *bufferPool
 
 // ==========
 // Error types.
@@ -129,7 +133,15 @@ func (vs *volumeSet) Set(value string) error {
        if _, err := os.Stat(value); err != nil {
                return err
        }
-       *vs = append(*vs, MakeUnixVolume(value, flagSerializeIO, flagReadonly))
+       var locker sync.Locker
+       if flagSerializeIO {
+               locker = &sync.Mutex{}
+       }
+       *vs = append(*vs, &UnixVolume{
+               root:     value,
+               locker:   locker,
+               readonly: flagReadonly,
+       })
        return nil
 }
 
@@ -195,7 +207,8 @@ func (vs *volumeSet) Discover() int {
 // permission arguments).
 
 func main() {
-       log.Println("Keep started: pid", os.Getpid())
+       log.Println("keepstore starting, pid", os.Getpid())
+       defer log.Println("keepstore exiting, pid", os.Getpid())
 
        var (
                data_manager_token_file string
@@ -224,7 +237,7 @@ func main() {
        flag.BoolVar(
                &never_delete,
                "never-delete",
-               false,
+               true,
                "If set, nothing will be deleted. HTTP 405 will be returned "+
                        "for valid DELETE requests.")
        flag.StringVar(
@@ -271,10 +284,49 @@ func main() {
                &pidfile,
                "pid",
                "",
-               "Path to write pid file")
+               "Path to write pid file during startup. This file is kept open and locked with LOCK_EX until keepstore exits, so `fuser -k pidfile` is one way to shut down. Exit immediately if there is an error opening, locking, or writing the pid file.")
+       flag.IntVar(
+               &maxBuffers,
+               "max-buffers",
+               maxBuffers,
+               fmt.Sprintf("Maximum RAM to use for data buffers, given in multiples of block size (%d MiB). When this limit is reached, HTTP requests requiring buffers (like GET and PUT) will wait for buffer space to be released.", BLOCKSIZE>>20))
 
        flag.Parse()
 
+       if never_delete != true {
+               log.Fatal("never_delete must be true, see #6221")
+       }
+
+       if maxBuffers < 0 {
+               log.Fatal("-max-buffers must be greater than zero.")
+       }
+       bufs = newBufferPool(maxBuffers, BLOCKSIZE)
+
+       if pidfile != "" {
+               f, err := os.OpenFile(pidfile, os.O_RDWR|os.O_CREATE, 0777)
+               if err != nil {
+                       log.Fatalf("open pidfile (%s): %s", pidfile, err)
+               }
+               err = syscall.Flock(int(f.Fd()), syscall.LOCK_EX|syscall.LOCK_NB)
+               if err != nil {
+                       log.Fatalf("flock pidfile (%s): %s", pidfile, err)
+               }
+               err = f.Truncate(0)
+               if err != nil {
+                       log.Fatalf("truncate pidfile (%s): %s", pidfile, err)
+               }
+               _, err = fmt.Fprint(f, os.Getpid())
+               if err != nil {
+                       log.Fatalf("write pidfile (%s): %s", pidfile, err)
+               }
+               err = f.Sync()
+               if err != nil {
+                       log.Fatalf("sync pidfile (%s): %s", pidfile, err)
+               }
+               defer f.Close()
+               defer os.Remove(pidfile)
+       }
+
        if len(volumes) == 0 {
                if volumes.Discover() == 0 {
                        log.Fatal("No volumes found.")
@@ -312,7 +364,7 @@ func main() {
                        log.Println("Running without a PermissionSecret. Block locators " +
                                "returned by this server will not be signed, and will be rejected " +
                                "by a server that enforces permissions.")
-                       log.Println("To fix this, use the -permission-key-file flag " +
+                       log.Println("To fix this, use the -blob-signing-key-file flag " +
                                "to specify the file containing the permission key.")
                }
        }
@@ -357,24 +409,9 @@ func main() {
                listener.Close()
        }(term)
        signal.Notify(term, syscall.SIGTERM)
+       signal.Notify(term, syscall.SIGINT)
 
-       if pidfile != "" {
-               f, err := os.Create(pidfile)
-               if err == nil {
-                       fmt.Fprint(f, os.Getpid())
-                       f.Close()
-               } else {
-                       log.Printf("Error writing pid file (%s): %s", pidfile, err.Error())
-               }
-       }
-
-       // Start listening for requests.
+       log.Println("listening at", listen)
        srv := &http.Server{Addr: listen}
        srv.Serve(listener)
-
-       log.Println("shutting down")
-
-       if pidfile != "" {
-               os.Remove(pidfile)
-       }
 }