package command import ( "fmt" "sort" "strings" "github.com/hashicorp/nomad/api" "github.com/hashicorp/nomad/api/contexts" "github.com/posener/complete" ) type EvalStatusCommand struct { Meta } func (c *EvalStatusCommand) Help() string { helpText := ` Usage: nomad eval status [options] Display information about evaluations. This command can be used to inspect the current status of an evaluation as well as determine the reason an evaluation did not place all allocations. General Options: ` + generalOptionsUsage() + ` Eval Status Options: -monitor Monitor an outstanding evaluation -verbose Show full information. -json Output the evaluation in its JSON format. -t Format and display evaluation using a Go template. ` return strings.TrimSpace(helpText) } func (c *EvalStatusCommand) Synopsis() string { return "Display evaluation status and placement failure reasons" } func (c *EvalStatusCommand) AutocompleteFlags() complete.Flags { return mergeAutocompleteFlags(c.Meta.AutocompleteFlags(FlagSetClient), complete.Flags{ "-json": complete.PredictNothing, "-monitor": complete.PredictNothing, "-t": complete.PredictAnything, "-verbose": complete.PredictNothing, }) } func (c *EvalStatusCommand) AutocompleteArgs() complete.Predictor { return complete.PredictFunc(func(a complete.Args) []string { client, err := c.Meta.Client() if err != nil { return nil } if err != nil { return nil } resp, _, err := client.Search().PrefixSearch(a.Last, contexts.Evals, nil) if err != nil { return []string{} } return resp.Matches[contexts.Evals] }) } func (c *EvalStatusCommand) Run(args []string) int { var monitor, verbose, json bool var tmpl string flags := c.Meta.FlagSet("eval status", FlagSetClient) flags.Usage = func() { c.Ui.Output(c.Help()) } flags.BoolVar(&monitor, "monitor", false, "") flags.BoolVar(&verbose, "verbose", false, "") flags.BoolVar(&json, "json", false, "") flags.StringVar(&tmpl, "t", "", "") if err := flags.Parse(args); err != nil { return 1 } // Check that we got exactly one evaluation ID args = flags.Args() // Get the HTTP client client, err := c.Meta.Client() if err != nil { c.Ui.Error(fmt.Sprintf("Error initializing client: %s", err)) return 1 } // If args not specified but output format is specified, format and output the evaluations data list if len(args) == 0 && json || len(tmpl) > 0 { evals, _, err := client.Evaluations().List(nil) if err != nil { c.Ui.Error(fmt.Sprintf("Error querying evaluations: %v", err)) return 1 } out, err := Format(json, tmpl, evals) if err != nil { c.Ui.Error(err.Error()) return 1 } c.Ui.Output(out) return 0 } if len(args) != 1 { c.Ui.Error(c.Help()) return 1 } evalID := args[0] // Truncate the id unless full length is requested length := shortId if verbose { length = fullId } // Query the allocation info if len(evalID) == 1 { c.Ui.Error(fmt.Sprintf("Identifier must contain at least two characters.")) return 1 } evalID = sanitizeUUIDPrefix(evalID) evals, _, err := client.Evaluations().PrefixList(evalID) if err != nil { c.Ui.Error(fmt.Sprintf("Error querying evaluation: %v", err)) return 1 } if len(evals) == 0 { c.Ui.Error(fmt.Sprintf("No evaluation(s) with prefix or id %q found", evalID)) return 1 } if len(evals) > 1 { // Format the evals out := make([]string, len(evals)+1) out[0] = "ID|Priority|Triggered By|Status|Placement Failures" for i, eval := range evals { failures, _ := evalFailureStatus(eval) out[i+1] = fmt.Sprintf("%s|%d|%s|%s|%s", limit(eval.ID, length), eval.Priority, eval.TriggeredBy, eval.Status, failures, ) } c.Ui.Error(fmt.Sprintf("Prefix matched multiple evaluations\n\n%s", formatList(out))) return 1 } // If we are in monitor mode, monitor and exit if monitor { mon := newMonitor(c.Ui, client, length) return mon.monitor(evals[0].ID, true) } // Prefix lookup matched a single evaluation eval, _, err := client.Evaluations().Info(evals[0].ID, nil) if err != nil { c.Ui.Error(fmt.Sprintf("Error querying evaluation: %s", err)) return 1 } // If output format is specified, format and output the data if json || len(tmpl) > 0 { out, err := Format(json, tmpl, eval) if err != nil { c.Ui.Error(err.Error()) return 1 } c.Ui.Output(out) return 0 } failureString, failures := evalFailureStatus(eval) triggerNoun, triggerSubj := getTriggerDetails(eval) statusDesc := eval.StatusDescription if statusDesc == "" { statusDesc = eval.Status } // Format the evaluation data basic := []string{ fmt.Sprintf("ID|%s", limit(eval.ID, length)), fmt.Sprintf("Status|%s", eval.Status), fmt.Sprintf("Status Description|%s", statusDesc), fmt.Sprintf("Type|%s", eval.Type), fmt.Sprintf("TriggeredBy|%s", eval.TriggeredBy), fmt.Sprintf("%s|%s", triggerNoun, triggerSubj), fmt.Sprintf("Priority|%d", eval.Priority), fmt.Sprintf("Placement Failures|%s", failureString), } if verbose { // NextEval, PreviousEval, BlockedEval basic = append(basic, fmt.Sprintf("Previous Eval|%s", eval.PreviousEval), fmt.Sprintf("Next Eval|%s", eval.NextEval), fmt.Sprintf("Blocked Eval|%s", eval.BlockedEval)) } c.Ui.Output(formatKV(basic)) if failures { c.Ui.Output(c.Colorize().Color("\n[bold]Failed Placements[reset]")) sorted := sortedTaskGroupFromMetrics(eval.FailedTGAllocs) for _, tg := range sorted { metrics := eval.FailedTGAllocs[tg] noun := "allocation" if metrics.CoalescedFailures > 0 { noun += "s" } c.Ui.Output(fmt.Sprintf("Task Group %q (failed to place %d %s):", tg, metrics.CoalescedFailures+1, noun)) c.Ui.Output(formatAllocMetrics(metrics, false, " ")) c.Ui.Output("") } if eval.BlockedEval != "" { c.Ui.Output(fmt.Sprintf("Evaluation %q waiting for additional capacity to place remainder", limit(eval.BlockedEval, length))) } } return 0 } func sortedTaskGroupFromMetrics(groups map[string]*api.AllocationMetric) []string { tgs := make([]string, 0, len(groups)) for tg := range groups { tgs = append(tgs, tg) } sort.Strings(tgs) return tgs } func getTriggerDetails(eval *api.Evaluation) (noun, subject string) { switch eval.TriggeredBy { case "job-register", "job-deregister", "periodic-job", "rolling-update", "deployment-watcher": return "Job ID", eval.JobID case "node-update": return "Node ID", eval.NodeID case "max-plan-attempts": return "Previous Eval", eval.PreviousEval default: return "", "" } }