planet: also sync to s3 changeset state config file

[chef.git] / cookbooks / planet / files / default / replication-bin / replicate-changesets

diff --git a/cookbooks/planet/files/default/replication-bin/replicate-changesets b/cookbooks/planet/files/default/replication-bin/replicate-changesets
old mode 100644 (file)
new mode 100755 (executable)
index 455fcbc..f6622e9
--- a/cookbooks/planet/files/default/replication-bin/replicate-changesets
+++ b/cookbooks/planet/files/default/replication-bin/replicate-changesets
@@ -100,7 +100,7 @@ class ChangesetBuilder
     xml["comments_count"] = res.num_tuples.to_s
 
     # early return if there aren't any comments
-    return unless res.num_tuples > 0
+    return unless res.num_tuples.positive?
 
     discussion = XML::Node.new("discussion")
     res.each do |row|
@@ -128,10 +128,7 @@ end
 # to guarantee sync for newly created directories.
 def fdirsync(d)
   while d != "/"
-    Dir.open(d) do |dh|
-      io = IO.for_fd(dh.fileno)
-      io.fsync
-    end
+    fsync(d)
     d = File.dirname(d)
   end
 end
@@ -141,9 +138,9 @@ end
 # replicated to a file.
 class Replicator
   def initialize(config)
-    @config = YAML.load(File.read(config))
-    @state = YAML.load(File.read(@config["state_file"]))
-    @conn = PGconn.connect(@config["db"])
+    @config = YAML.safe_load(File.read(config))
+    @state = YAML.safe_load(File.read(@config["state_file"]), [Time])
+    @conn = PG::Connection.connect(@config["db"])
     # get current time from the database rather than the current system
     @now = @conn.exec("select now() as now").map { |row| Time.parse(row["now"]) }[0]
   end
@@ -167,7 +164,7 @@ class Replicator
     new_ids = @conn
               .exec("select distinct changeset_id from changeset_comments where created_at >= '#{last_run}' and created_at < '#{@now}' and visible")
               .map { |row| row["changeset_id"].to_i }
-              .select { |c_id| !cs_ids.include?(c_id) }
+              .reject { |c_id| cs_ids.include?(c_id) }
 
     new_ids.each do |id|
       @conn
@@ -187,8 +184,8 @@ class Replicator
     { "version" => "0.6",
       "generator" => "replicate_changesets.rb",
       "copyright" => "OpenStreetMap and contributors",
-      "attribution" => "http://www.openstreetmap.org/copyright",
-      "license" => "http://opendatacommons.org/licenses/odbl/1-0/" }
+      "attribution" => "https://www.openstreetmap.org/copyright",
+      "license" => "https://opendatacommons.org/licenses/odbl/1-0/" }
       .each { |k, v| doc.root[k] = v }
 
     builder = ChangesetBuilder.new(@now, @conn)
@@ -207,6 +204,10 @@ class Replicator
     @config["data_dir"] + format("/%03d/%03d/%03d", sequence / 1000000, (sequence / 1000) % 1000, (sequence % 1000))
   end
 
+  def s3_stem
+    @config["s3_dir"] + format("/%03d/%03d/%03d", sequence / 1000000, (sequence / 1000) % 1000, (sequence % 1000))
+  end
+
   def write_tmp_files!(changesets)
     data_file = data_stem + ".osm.gz"
     tmp_state = @config["state_file"] + ".tmp"
@@ -216,9 +217,7 @@ class Replicator
     Zlib::GzipWriter.open(tmp_data) do |fh|
       fh.write(changeset_dump(changesets))
     end
-    File.open(tmp_state, "w") do |fh|
-      fh.write(YAML.dump(@state))
-    end
+    File.write(tmp_state, YAML.dump(@state))
 
     # fsync the files in their old locations.
     fsync(tmp_data)
@@ -257,29 +256,43 @@ class Replicator
     # from the dirent and has been updated to account for any allocations.
     fdirsync(File.dirname(data_file))
     fdirsync(File.dirname(@config["state_file"]))
+
+    if @config["s3_dir"]
+      s3_file = s3_stem + ".osm.gz"
+      s3_state_config_file = @config["s3_dir"] + "/state.yaml"
+      s3_state_file = s3_stem + ".state.txt"
+
+      system("/opt/awscli/v2/current/bin/aws", "--profile=osm-pds-upload", "s3", "cp", "--storage-class=INTELLIGENT_TIERING", "--no-progress", data_file, s3_file)
+      system("/opt/awscli/v2/current/bin/aws", "--profile=osm-pds-upload", "s3", "cp", "--storage-class=INTELLIGENT_TIERING", "--no-progress", @config["state_file"], s3_state_config_file)
+      system("/opt/awscli/v2/current/bin/aws", "--profile=osm-pds-upload", "s3", "cp", "--storage-class=INTELLIGENT_TIERING", "--no-progress", data_state_file, s3_state_file)
+    end
   end
 
   # saves new state (including the changeset dump xml)
   def save!
-    File.open(@config["state_file"], "r") do |fl|
-      fl.flock(File::LOCK_EX)
+    File.open(@config["lock_file"], File::RDWR | File::CREAT, 0o600) do |fl|
+      # take the lock in non-blocking mode. if this process doesn't get the lock
+      # then another will be run from cron shortly. this prevents a whole bunch
+      # of processes queueing on the lock and causing weirdness if/when they
+      # get woken up in a random order.
+      got_lock = fl.flock(File::LOCK_EX | File::LOCK_NB)
+      break unless got_lock
 
       # try and write the files to tmp locations and then
       # move them into place later, to avoid in-progress
       # clashes, or people seeing incomplete files.
       begin
+        changesets = open_changesets
         @state["sequence"] = sequence
         @state["last_run"] = @now
-        changesets = open_changesets
 
         write_tmp_files!(changesets)
 
         move_tmp_files_into_place!
 
         fl.flock(File::LOCK_UN)
-
-      rescue
-        STDERR.puts("Error! Couldn't update state.")
+      rescue StandardError
+        warn "Error! Couldn't update state."
         fl.flock(File::LOCK_UN)
         raise
       end
@@ -291,6 +304,9 @@ begin
   rep = Replicator.new(ARGV[0])
   rep.save!
 rescue StandardError => e
-  STDERR.puts "ERROR: #{e.message}"
+  warn "ERROR: #{e.message}"
+  e.backtrace.each do |frame|
+    warn "ERROR: #{frame}"
+  end
   exit 1
 end