"rescue Exception" not rescuing Timeout::Error in net_http - ruby

Question

We appear to have a situation where rescue Exception is not catching a particular exception.
I'm trying to send an email alert about any exception that occurs, and then continue processing. We've put in the requisit handling of intentional exits. We want the loop to keep going, after alerting us, for anything else.
The exception that is not being caught is ostensibly Timeout::Error, according to the stack trace.
Here is the stack trace, with references to my intermediate code removed (the last line of my code is request.rb:93):
/opt/ruby-enterprise/lib/ruby/1.8/timeout.rb:64:in `rbuf_fill': execution expired (Timeout::Error)
from /opt/ruby-enterprise/lib/ruby/1.8/net/protocol.rb:134:in `rbuf_fill'
from /opt/ruby-enterprise/lib/ruby/1.8/net/protocol.rb:116:in `readuntil'
from /opt/ruby-enterprise/lib/ruby/1.8/net/protocol.rb:126:in `readline'
from /opt/ruby-enterprise/lib/ruby/1.8/net/http.rb:2028:in `read_status_line'
from /opt/ruby-enterprise/lib/ruby/1.8/net/http.rb:2017:in `read_new'
from /opt/ruby-enterprise/lib/ruby/1.8/net/http.rb:1051:in `__request__'
from /mnt/data/blueleaf/releases/20150211222522/vendor/bundle/ruby/1.8/gems/rest-client-1.6.7/lib/restclient/net_http_ext.rb:51:in `request'
from /opt/ruby-enterprise/lib/ruby/1.8/net/http.rb:1037:in `__request__'
from /opt/ruby-enterprise/lib/ruby/1.8/net/http.rb:543:in `start'
from /opt/ruby-enterprise/lib/ruby/1.8/net/http.rb:1035:in `__request__'
from /mnt/data/blueleaf/releases/20150211222522/vendor/bundle/ruby/1.8/gems/rest-client-1.6.7/lib/restclient/net_http_ext.rb:51:in `request'
from /mnt/data/blueleaf/releases/20150211222522/app/models/dst/request.rb:93:in `send'
[intermediate code removed]
from script/dst_daemon.rb:49
from script/dst_daemon.rb:46:in `each'
from script/dst_daemon.rb:46
from /opt/ruby-enterprise/lib/ruby/1.8/benchmark.rb:293:in `measure'
from script/dst_daemon.rb:45
from script/dst_daemon.rb:24:in `loop'
from script/dst_daemon.rb:24
from script/runner:3:in `eval'
from /mnt/data/blueleaf/releases/20150211222522/vendor/bundle/ruby/1.8/gems/rails-2.3.14/lib/commands/runner.rb:46
from script/runner:3:in `require'
Here is request.rb#send, with line 93 indicated with a comment:
def send
build
uri = URI.parse([DST::Request.configuration[:prefix], #path].join('/'))
https = Net::HTTP.new(uri.host, uri.port)
https.use_ssl = true
https.verify_mode = OpenSSL::SSL::VERIFY_NONE
https_request = Net::HTTP::Post.new(uri.request_uri.tap{|e| debug_puts "\nURL: #{e}, host:#{uri.host}"})
# line 93:
https_request.body = request
response = https.request(https_request)
# the rest should be irrelevant
Here is dst_daemon.rb; line 49 is indicated with a comment, and the rescue Exception that should catch anything other than deliberate interrupts is near the end:
DST::Request.environment = :production
class DST::Request::RequestFailed < Exception; end
Thread.abort_on_exception = true
SEMAPHORE = 'import/dst/start.txt' unless defined?(SEMAPHORE)
DEBUG_DST = 'import/dst/debug.txt' unless defined?(DEBUG_DST)
DEBUG_LOG = 'import/dst/debug.log' unless defined?(DEBUG_LOG)
def debug_dst(*args)
File.open(DEBUG_LOG, 'a') do |f|
f.print "#{Time.now.localtime}: "
f.puts(*args)
end if debug_dst?
end
def debug_dst?
File.exist?(DEBUG_DST)
end
dst_ids = [Institution::BAA_DST_WS_CLIENT_ID, Institution::BAA_DST_WS_DEALER_ID]
institutions = Institution.find_all_by_baa_api_financial_institution_id(dst_ids)
DST::Collector.prime_key!
loop do
begin
if File.exist?(SEMAPHORE)
debug_dst 'waking up...'
custodians = InstitutionAccount.acts_as_baa_custodian.
find_all_by_institution_id(institutions).select(&:direct?)
good,bad = custodians.partition do |c|
c.custodian_users.map{|e2|e2.custodian_passwords.count(:conditions => ['expired is not true']) == 1}.all?
end
if bad.present?
msg = " skipping: \n"
bad.each do |c|
msg += " #{c.user.full_name_or_email}, custodian id #{c.id}: "
c.custodian_users.each{|cu| msg += "#{cu.username}:#{cu.custodian_passwords.count(:conditions => ['expired is not true'])}; "}
msg += "\n"
end
AdminSimpleMailer.deliver_generic_mail("DST Daemon skipping #{bad.size} connections", msg)
debug_dst msg
end
Benchmark.measure do
good.each do |custodian|
begin
debug_dst " collecting for: #{custodian.name}, #{custodian.subtitle}, (#{custodian.id.inspect})"
# line 49:
DST::Collector.new(custodian, 0).collect!
rescue DST::Request::PasswordFailed, DST::Request::RequestFailed => e
message = e.message + "\n\n" + e.backtrace.join("\n")
AdminSimpleMailer.deliver_generic_mail("DST Daemon Connection Failed #{e.class.name}", message)
debug_dst " skipping, #{e.class}"
end
end
end.tap{|duration| debug_dst "collection done, duration #{duration.real.to_f/60} minutes. importing" }
DST::Strategy.new(Date.yesterday, :recompute => true).import!
debug_dst 'import done.'
rm SEMAPHORE, :verbose => debug_dst?
else
debug_dst 'sleeping.' if Time.now.strftime("%M").to_i % 5 == 0
end
rescue SystemExit, Interrupt
raise
rescue Exception => e
message = e.message + "\n\n" + e.backtrace.join("\n")
AdminSimpleMailer.deliver_generic_mail("DST Daemon Exception #{e.class.name}", message)
ensure
sleep 60
end
end
Shouldn't it be impossible for this loop to exit with a stack trace other than from SystemExit or Interrupt?

Answer 1

As you probably know already, calling raise inside a rescue block will raise the exception to the caller.
Since Timeout::Error is an Interrupt in ruby 1.8*, the timeout exception raised by net_http gets handled in the rescue SystemExit, Interrupt block rather than in the following rescue Exception => e.
To verify that Timeout::Error is an Interrupt, just evaluate Timeout::Error.ancestors. What you get out of that is the hierarchy of classes Timeout::Error inherits from.
*this is no longer the case in ruby1.9.