def add_comments(xml, cs)
# grab the visible changeset comments as well
- res = @conn.exec("select cc.author_id, u.display_name as author, cc.body, cc.created_at from changeset_comments cc join users u on cc.author_id=u.id where cc.changeset_id=#{cs.id} and cc.visible order by cc.created_at asc")
+ res = @conn.exec("select cc.id, cc.author_id, u.display_name as author, cc.body, (cc.created_at at time zone 'utc') as created_at from changeset_comments cc join users u on cc.author_id=u.id where cc.changeset_id=#{cs.id} and cc.visible order by cc.created_at asc")
xml["comments_count"] = res.num_tuples.to_s
# early return if there aren't any comments
discussion = XML::Node.new("discussion")
res.each do |row|
comment = XML::Node.new("comment")
+ comment["id"] = row["id"]
comment["uid"] = row["author_id"]
comment["user"] = xml_sanitize(row["author"])
comment["date"] = Time.parse(row["created_at"]).getutc.xmlschema
# sync a directory to guarantee it's on disk. have to recurse to the root
# 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
+ while d != "/" && d != "."
+ fsync(d)
d = File.dirname(d)
end
end
class Replicator
def initialize(config)
@config = YAML.safe_load(File.read(config))
- @state = YAML.safe_load(File.read(@config["state_file"]), [Time])
- @conn = PGconn.connect(@config["db"])
+ @state = YAML.safe_load(File.read(@config["state_file"]), :permitted_classes => [Time], :fallback => {})
+ @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
# for us to look at anything that was closed recently, and filter from
# there.
changesets = @conn
- .exec("select id, created_at, closed_at, num_changes from changesets where closed_at > ((now() at time zone 'utc') - '1 hour'::interval)")
+ .exec("select id, (created_at at time zone 'utc') as created_at, (closed_at at time zone 'utc') as closed_at, num_changes from changesets where (closed_at at time zone 'utc') > ((now() at time zone 'utc') - '1 hour'::interval)")
.map { |row| Changeset.new(row) }
.select { |cs| cs.activity_between?(last_run, @now) }
# but also add any changesets which have new comments
new_ids = @conn
- .exec("select distinct changeset_id from changeset_comments where created_at >= '#{last_run}' and created_at < '#{@now}' and visible")
+ .exec("select distinct changeset_id from changeset_comments where (created_at at time zone 'utc') >= '#{last_run}' and (created_at at time zone 'utc') < '#{@now}' and visible")
.map { |row| row["changeset_id"].to_i }
.reject { |c_id| cs_ids.include?(c_id) }
new_ids.each do |id|
@conn
- .exec("select id, created_at, closed_at, num_changes from changesets where id=#{id}")
+ .exec("select id, (created_at at time zone 'utc') as created_at, (closed_at at time zone 'utc') as closed_at, num_changes from changesets where id=#{id}")
.map { |row| Changeset.new(row) }
.each { |cs| changesets << cs }
end
@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"
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)
# 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_file = s3_stem + ".state.txt"
+ s3_state_config_file = @config["s3_dir"] + "/state.yaml"
+
+ 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", data_state_file, s3_state_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)
+ 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
fl.flock(File::LOCK_UN)
rescue StandardError
- STDERR.puts("Error! Couldn't update state.")
+ warn "Error! Couldn't update state."
fl.flock(File::LOCK_UN)
raise
end
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