path: root/guix-build-coordinator/agent.scm

;;; Guix Build Coordinator
;;;
;;; Copyright © 2020 Christopher Baines <mail@cbaines.net>
;;;
;;; This file is part of the guix-build-coordinator.
;;;
;;; The Guix Build Coordinator is free software; you can redistribute
;;; it and/or modify it under the terms of the GNU General Public
;;; License as published by the Free Software Foundation; either
;;; version 3 of the License, or (at your option) any later version.
;;;
;;; The Guix Build Coordinator is distributed in the hope that it will
;;; be useful, but WITHOUT ANY WARRANTY; without even the implied
;;; warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
;;; See the GNU General Public License for more details.
;;;
;;; You should have received a copy of the GNU General Public License
;;; along with the guix-data-service.  If not, see
;;; <http://www.gnu.org/licenses/>.

(define-module (guix-build-coordinator agent)
  #:use-module (srfi srfi-1)
  #:use-module (srfi srfi-11)
  #:use-module (ice-9 match)
  #:use-module (ice-9 futures)
  #:use-module (ice-9 atomic)
  #:use-module (ice-9 threads)
  #:use-module (ice-9 exceptions)
  #:use-module (rnrs bytevectors)
  #:use-module (web http)
  #:use-module (guix store)
  #:use-module (guix derivations)
  #:use-module (guix base32)
  #:use-module (guix-build-coordinator utils)
  #:use-module (guix-build-coordinator agent-messaging http)
  #:export (run-agent))

(define (run-agent uuid coordinator-uri password
                   systems
                   max-parallel-builds
                   derivation-substitute-urls
                   non-derivation-substitute-urls)
  (define (fetch-new-jobs count)
    (let ((received-builds
           (fetch-builds-for-agent coordinator-uri uuid password
                                   systems
                                   #:count count)))
      (simple-format #t "requested ~A builds, received ~A\n"
                     count (length received-builds))
      received-builds))

  (define (process-job build)
    (let ((build-id (assoc-ref build "uuid"))
          (derivation-name (assoc-ref build "derivation-name")))
      (simple-format #t "~A: setting up to build: ~A\n"
                     build-id derivation-name)
      (with-store store
        (let ((pre-build-status (pre-build-process
                                 store
                                 derivation-substitute-urls
                                 non-derivation-substitute-urls
                                 derivation-name)))
          (if (eq? (assq-ref pre-build-status 'result) 'success)
              (begin
                (simple-format #t "~A: setup successful, building: ~A\n"
                               build-id derivation-name)
                (report-build-start coordinator-uri uuid password
                                    build-id)
                (let* ((result (perform-build store derivation-name))
                       ;; TODO Check this handles timezones right
                       (end-time (localtime (current-time) "UTC")))

                  (agent-submit-log-file uuid coordinator-uri password
                                         build-id derivation-name)

                  ((if result
                       post-build-success
                       post-build-failure)
                   uuid coordinator-uri password
                   build-id
                   derivation-name
                   end-time)))
              (begin
                (simple-format #t "~A: failure: ~A\n" build-id pre-build-status)
                (report-setup-failure coordinator-uri uuid password
                                      build-id
                                      pre-build-status)))))))

  (let-values (((process-job-with-queue count-jobs)
                (create-work-queue max-parallel-builds
                                   process-job)))
    (let ((details (submit-status coordinator-uri uuid password 'idle)))
      (let ((builds (vector->list (assoc-ref details "builds"))))
        (for-each
         (lambda (job-args)
           (process-job-with-queue job-args))
         builds)))

    (while #t
      (let ((job-count (count-jobs)))
        (when (< job-count max-parallel-builds)
          (match (fetch-new-jobs (- max-parallel-builds job-count))
            (()
             ;; No new jobs available
             (sleep 30))
            ((jobs ...)
             (for-each
              (lambda (job-args)
                (process-job-with-queue job-args))
              jobs)))))
      (sleep 5))))

(define (agent-submit-log-file uuid coordinator-uri password
                               build-id derivation-name)
  (retry-on-error
   (lambda ()
     (let ((log-file (derivation-log-file derivation-name)))
       (unless log-file
         (raise-exception
          (make-exception-with-message
           (simple-format #f "log file missing for ~A (~A)"
                          derivation-name build-id))))

       (simple-format #t "~A: uploading log file ~A\n"
                      build-id log-file)
       (submit-log-file
        coordinator-uri uuid password
        build-id
        log-file)))
   #:times 6
   #:delay 30))

(define (pre-build-process store
                           derivation-substitute-urls
                           non-derivation-substitute-urls
                           derivation-name)
  (define (find-missing-inputs derivation inputs)
    (let* ((output-paths
            (append-map derivation-input-output-paths inputs))
           (missing-paths
            (remove (lambda (path)
                      (valid-path? store path))
                    output-paths))
           (path-substitutes
            (begin
              (set-build-options store #:substitute-urls
                                 non-derivation-substitute-urls)

              (map (lambda (file)
                     (and
                      (has-substiutes-no-cache? non-derivation-substitute-urls
                                                file)
                      (if (has-substitutes? store file)
                          #t
                          (begin
                            (simple-format
                             #t "warning: a substitute should be available for ~A, but the daemon claims it's not\n"
                             file)
                            #f))))
                   missing-paths))))

      (if (null? missing-paths)
          '()
          (if (member #f path-substitutes)
              (fold (lambda (file substitute-available? result)
                      (if substitute-available?
                          result
                          (cons file result)))
                    '()
                    missing-paths
                    path-substitutes)
              (begin
                (retry-on-error
                 (lambda ()
                   (with-timeout (* 1000 60 10) ; 10 minutes
                     (raise-exception
                      (make-exception-with-message "timeout fetching inputs"))
                     (begin
                       ;; Download the substitutes
                       (set-build-options store
                                          #:substitute-urls
                                          non-derivation-substitute-urls)

                       (build-things store missing-paths))))
                 #:times 6
                 #:delay 60)

                ;; Double check everything is actually present.
                (let ((missing-files (remove (lambda (path)
                                               (valid-path? store path))
                                             output-paths)))
                  (if (null? missing-files)
                      '()
                      (begin
                        (simple-format
                         (current-error-port)
                         "warning: failed to fetch substitutes for: ~A\n"
                         missing-files)

                        (let ((unavailable-outputs
                               (delete-duplicates
                                (append-map
                                 (lambda (missing-output)
                                   (find-missing-substitutes-for-output
                                    non-derivation-substitute-urls
                                    missing-output))
                                 missing-files))))

                          (simple-format
                           (current-error-port)
                           "warning: the following outputs are missing:\n~A\n"
                           (string-join
                            (map (lambda (output)
                                   (string-append
                                    " - " output))
                                 unavailable-outputs)))

                          unavailable-outputs)))))))))

  (define (delete-outputs derivation)
    (let* ((outputs (derivation-outputs derivation))
           (output-file-names
            (map derivation-output-path (map cdr outputs))))
      (if (any file-exists? output-file-names)
          (catch
            #t
            (lambda ()
              ;; There can be issues deleting links when collecting garbage
              ;; from multiple threads
              (monitor
               (with-store store
                 (delete-paths store output-file-names)))
              #t)
            (lambda (key args)
              (simple-format (current-error-port)
                             "error: delete-outputs: ~A ~A\n"
                             key args)
              #f))
          #t)))

  (let ((derivation
         (if (valid-path? store derivation-name)
             (read-derivation-from-file derivation-name)
             (begin (retry-on-error
                     (lambda ()
                       (substitute-derivation derivation-name
                                              #:substitute-urls
                                              derivation-substitute-urls))
                     #:times 12
                     #:delay 20)
                    (read-derivation-from-file derivation-name)))))
    (match (delete-outputs derivation)
      (#t
       (let ((missing-inputs
              (find-missing-inputs derivation (derivation-inputs derivation))))
         (if (null? missing-inputs)
             '((result . success))
             `((result         . failure)
               (failure_reason . missing_inputs)
               (missing_inputs . ,(list->vector missing-inputs))))))
      (failure
       '((result         . failure)
         (failure_reason . could_not_delete_outputs))))))

(define (perform-build store derivation-name)
  (set-build-options store
                     #:use-substitutes? #f)

  (parameterize ((current-build-output-port (%make-void-port "w")))
    (catch #t
      (lambda ()
        (build-things store (list derivation-name))
        #t)
      (lambda (key . args)
        (simple-format (current-error-port)
                       "error: build: ~A ~A\n"
                       key args)

        #f))))

(define (post-build-failure uuid coordinator-uri password
                            build-id derivation end-time)
  (simple-format #t "build ~A failed, reporting to coordinator\n"
                 build-id)
  (submit-build-result coordinator-uri uuid password
                       build-id
                       `((result   . failure)
                         (end_time . ,(strftime "%F %T" end-time)))))

(define (post-build-success uuid coordinator-uri password
                            build-id derivation end-time)
  (define output-details
    (map
     (match-lambda
       ((output-name . output)
        (let ((path-info (with-store store
                           (query-path-info
                            store
                            (derivation-output-path output)))))
          `((name       . ,output-name)
            (hash       . ,(bytevector->nix-base32-string
                            (path-info-hash path-info)))
            (size       . ,(path-info-nar-size path-info))
            (references . ,(list->vector
                            (map basename
                                 (path-info-references path-info))))))))

     (derivation-outputs (read-derivation-from-file derivation))))

  (define (attempt-submit-build-result)
    (with-exception-handler
      (lambda (exn)
        (unless (agent-error-from-coordinator? exn)
          (raise-exception exn))

        (let ((details
               (agent-error-from-coordinator-details exn)))
          (if (string? details)
              (cond
               ((string=? details "build_already_processed")
                (simple-format
                 #t "warning: coordinator indicates this build is already marked as processed\n")
                #t)
               ((string=? details "missing_build_log_file")
                ;; Retry submitting the log file
                (agent-submit-log-file uuid coordinator-uri password
                                       build-id derivation)

                (attempt-submit-build-result))
               (else
                (raise-exception
                 (make-exception
                  (make-exception-with-message
                   "unrecognised error from coordinator")
                  (make-exception-with-irritants
                   details)))))
              (or
               (and=> (assoc-ref details "missing_output")
                      (lambda (missing-output-name)
                        (let ((missing-output
                               (any (match-lambda
                                      ((name . output)
                                       (if (string=? name missing-output-name)
                                           output
                                           #f)))
                                    (derivation-outputs
                                     (read-derivation-from-file derivation)))))
                          (unless missing-output
                            (raise-exception
                             (make-exception
                              (make-exception-with-message
                               "unknown missing output")
                              (make-exception-with-irritants
                               missing-output-name))))

                          (submit-one-output missing-output-name
                                             missing-output))

                        (attempt-submit-build-result)))
               (raise-exception
                (make-exception
                 (make-exception-with-message
                  "unrecognised error from coordinator")
                 (make-exception-with-irritants
                  details)))))))
      (lambda ()
        (submit-build-result
         coordinator-uri uuid password build-id
         `((result   . success)
           (end_time . ,(strftime "%F %T" end-time))
           (outputs  . ,(list->vector output-details)))))
      #:unwind? #t))

  (define (submit-one-output output-name output)
    (simple-format #t "submitting output ~A\n"
                   (derivation-output-path output))
    (submit-output coordinator-uri uuid password
                   build-id output-name
                   (derivation-output-path output)))

  (simple-format #t "build ~A successful, reporting to coordinator\n"
                 build-id)

  (for-each (match-lambda
              ((output-name . output)
               (submit-one-output output-name output)))
            (derivation-outputs (read-derivation-from-file derivation)))

  (simple-format #t "finished submitting outputs, reporting result\n")
  (attempt-submit-build-result))