Backfill excerpts for servers registered under http

app/models/log_excerpt.rb

@@ -12,7 +12,7 @@ class LogExcerpt < ApplicationRecord
   # (missing or empty objects) so that backfill can resume by max report_id.
   # Network errors are raised to the caller.
   def self.capture(report)
-    return nil unless %r{\Ahttps?://rubyci\.s3\.amazonaws\.com/}.match?(report.server&.uri.to_s)
+    return nil unless report.server&.rubyci_s3?
     content = fetch_content(report)
     excerpt = find_or_initialize_by(report_id: report.id)
     excerpt.content = content
@@ -35,14 +35,19 @@ def self.fetch_content(report)
     sections.map { |s| truncate_bytes(s, budget + slack) }.join("\n")
   end
 
-  # GET a gzipped text file. Returns nil on 404. S3 serves these either with
-  # Content-Encoding: gzip (Net::HTTP inflates the body) or as raw gzip bytes.
+  # GET a gzipped text file. Returns nil when the object cannot be read, which
+  # covers a plain 404 and an object Intelligent-Tiering has moved to an
+  # archive tier. Neither is retryable, so both become an empty excerpt rather
+  # than an error. S3 serves these either with Content-Encoding: gzip
+  # (Net::HTTP inflates the body) or as raw gzip bytes.
   def self.fetch_text(uri)
-    uri = URI(uri)
-    res = Net::HTTP.start(uri.host, uri.port, open_timeout: 10, read_timeout: 30, use_ssl: uri.scheme == "https") do |h|
+    # Servers are registered under both http:// and https://; always use TLS.
+    uri = URI(uri.to_s.sub(%r{\Ahttp://}, "https://"))
+    res = Net::HTTP.start(uri.host, uri.port, open_timeout: 10, read_timeout: 30, use_ssl: true) do |h|
       h.get(uri.path)
     end
     return nil if res.code == "404"
+    return nil if res.code == "403" && res.body.to_s.include?("InvalidObjectState")
     res.value
     body = res.body
     begin

app/models/report.rb

@@ -289,7 +289,7 @@ def self.fetch_recent
     ary = []
     Server.order(:id).all.each do |server|
       puts server.uri
-      if %r<\Ahttps?://rubyci.s3.amazonaws.com/>.match?(server.uri)
+      if server.rubyci_s3?
         t = Time.now
         n = self.get_reports_rubyci_s3(s3, server)
         ary << [server.uri, Time.now-t, n]

app/models/server.rb

@@ -3,6 +3,21 @@ class Server < ApplicationRecord
   validates :uri, :length => { :in => 20..200 }
   validates :ordinal, :numericality => true, :uniqueness => true
 
+  RUBYCI_S3_HOST = 'rubyci.s3.amazonaws.com'
+  RUBYCI_S3_URI = %r{\Ahttps?://#{Regexp.escape(RUBYCI_S3_HOST)}/}
+
+  # Most rows still carry the http:// form these URIs were first registered
+  # with, so both schemes have to match. Keep the Ruby and SQL forms together
+  # or they drift apart and each caller covers a different set of servers.
+  scope :rubyci_s3, -> {
+    where("uri LIKE :http OR uri LIKE :https",
+      http: "http://#{RUBYCI_S3_HOST}/%", https: "https://#{RUBYCI_S3_HOST}/%")
+  }
+
+  def rubyci_s3?
+    RUBYCI_S3_URI.match?(uri.to_s)
+  end
+
   def recent_uri(branch)
     "#{uri.chomp('/')}/ruby-#{branch}/recent.html"
   end

lib/tasks/log_excerpts.rake

@@ -26,6 +26,7 @@ end
 namespace :log_excerpts do
   desc "Backfill log excerpts from S3. FROM/TO (UTC date, default last 1 year), START_ID to force resume point, SLEEP seconds between reports (default 0.1), FORCE=1 to refetch reports that already have an excerpt"
   task :backfill => :environment do
+    $stdout.sync = true # detached dynos block-buffer, hiding progress until exit
     from = LogExcerptTaskEnv.time("FROM", 1.year.ago)
     to = LogExcerptTaskEnv.time("TO", Time.now)
     interval = LogExcerptTaskEnv.float("SLEEP", 0.1)
@@ -35,7 +36,7 @@ namespace :log_excerpts do
       where(datetime: from..to).
       where.not(ltsv: nil).
       where("ltsv NOT LIKE '%result:success%'").
-      where("servers.uri LIKE 'https://rubyci.s3.amazonaws.com/%'")
+      merge(Server.rubyci_s3)
 
     # Resume by skipping reports that already have an excerpt. Using the max
     # captured report_id instead would skip everything, because scan_recent_ltsv
@@ -72,6 +73,7 @@ namespace :log_excerpts do
 
   desc "Delete log excerpts whose report datetime is older than KEEP_DAYS (default 366) days. DRY_RUN=1 to only count."
   task :prune => :environment do
+    $stdout.sync = true
     keep_days = LogExcerptTaskEnv.integer("KEEP_DAYS", 366)
     abort "KEEP_DAYS must be at least 1, got #{keep_days}" if keep_days < 1
     cutoff = keep_days.days.ago

test/models/log_excerpt_test.rb

@@ -101,6 +101,15 @@ def create_report(server, attrs = {})
     assert_nil report.reload.log_excerpt
   end
 
+  test "capture handles servers registered under http" do
+    server = create_server("cap-http", uri: "http://rubyci.s3.amazonaws.com/cap-http/")
+    report = create_report(server)
+    stub_fetch_text(->(uri) { "body" }) do
+      LogExcerpt.capture(report)
+    end
+    assert_equal "body\nbody", report.reload.log_excerpt.content
+  end
+
   test "capture is idempotent per report" do
     server = create_server("cap-idem")
     report = create_report(server)

test/models/server_test.rb

@@ -0,0 +1,30 @@
+require "test_helper"
+
+class ServerTest < ActiveSupport::TestCase
+  @@ordinal = 500
+
+  def create_server(uri)
+    Server.create!(name: "srv-#{@@ordinal}", uri: uri, ordinal: (@@ordinal += 1))
+  end
+
+  test "rubyci_s3? accepts both schemes" do
+    assert_predicate create_server("https://rubyci.s3.amazonaws.com/ubuntu/"), :rubyci_s3?
+    assert_predicate create_server("http://rubyci.s3.amazonaws.com/rhel9"), :rubyci_s3?
+  end
+
+  test "rubyci_s3? rejects other hosts" do
+    assert_not create_server("https://example.com/rubyci.s3.amazonaws.com/x").rubyci_s3?
+    assert_not create_server("https://rubyci.s3.amazonaws.com.evil.test/x").rubyci_s3?
+    assert_not create_server("http://ci.rvm.jp/chkbuild/logs/").rubyci_s3?
+  end
+
+  test "rubyci_s3 scope selects the same servers as rubyci_s3?" do
+    servers = [
+      create_server("https://rubyci.s3.amazonaws.com/scope-https/"),
+      create_server("http://rubyci.s3.amazonaws.com/scope-http/"),
+      create_server("http://ci.rvm.jp/chkbuild/scope-other/"),
+    ]
+    selected = Server.rubyci_s3.where(id: servers.map(&:id)).order(:id).to_a
+    assert_equal servers.select(&:rubyci_s3?), selected
+  end
+end