Londiste: try to show exact event info on failure.

This should help debugging replication problems.

It does not work on first error, as we batch together a lot of SQL
statements, but on retry Londiste start replaying events one-by-one,
the the exact event can be spotted.

DBScript unnecessarily overrided emsg, drop that.

2 files changed, 21 insertions, 1 deletions

diff --git a/python/londiste/playback.py b/python/londiste/playback.py
index fddf3b5c..4129cbe1 100644
--- a/python/londiste/playback.py
+++ b/python/londiste/playback.py
@@ -300,6 +300,8 @@ class Replicator(CascadedWorker):
     copy_table_name = None # filled by Copytable()
     sql_list = []
 
+    current_event = None
+
     def __init__(self, args):
         """Replication init."""
         CascadedWorker.__init__(self, 'londiste3', 'db', args)
@@ -337,6 +339,8 @@ class Replicator(CascadedWorker):
     def process_remote_batch(self, src_db, tick_id, ev_list, dst_db):
         "All work for a batch.  Entry point from SetConsumer."
 
+        self.current_event = None
+
         # this part can play freely with transactions
 
         if not self.code_check_done:
@@ -592,7 +596,13 @@ class Replicator(CascadedWorker):
 
     def process_remote_event(self, src_curs, dst_curs, ev):
         """handle one event"""
+
         self.log.debug("New event: id=%s / type=%s / data=%s / extra1=%s" % (ev.id, ev.type, ev.data, ev.extra1))
+
+        # set current_event only if processing them one-by-one
+        if self.work_state < 0:
+            self.current_event = ev
+
         if ev.type in ('I', 'U', 'D'):
             self.handle_data_event(ev, dst_curs)
         elif ev.type[:2] in ('I:', 'U:', 'D:'):
@@ -618,6 +628,9 @@ class Replicator(CascadedWorker):
         else:
             CascadedWorker.process_remote_event(self, src_curs, dst_curs, ev)
 
+        # no point keeping it around longer
+        self.current_event = None
+
     def handle_data_event(self, ev, dst_curs):
         """handle one truncate event"""
         t = self.get_table_by_name(ev.extra1)
@@ -955,6 +968,14 @@ class Replicator(CascadedWorker):
                 return
         CascadedWorker.copy_event(self, dst_curs, ev, filtered_copy)
 
+    def exception_hook(self, det, emsg):
+        # add event info to error message
+        if self.current_event:
+            ev = self.current_event
+            info = "[ev_id=%d,ev_txid=%d] " % (ev.ev_id,ev.ev_txid)
+            emsg = info + emsg
+        super(Replicator, self).exception_hook(det, emsg)
+
 if __name__ == '__main__':
     script = Replicator(sys.argv[1:])
     script.start()
diff --git a/python/skytools/scripting.py b/python/skytools/scripting.py
index 049da251..ff8eac79 100644
--- a/python/skytools/scripting.py
+++ b/python/skytools/scripting.py
@@ -774,7 +774,6 @@ class DBScript(BaseScript):
             sql = getattr(curs, 'query', '?')
             if len(sql) > 200: # avoid logging londiste huge batched queries
                 sql = sql[:60] + " ..."
-            emsg = str(d).strip()
             self.log.exception("Job %s got error on connection '%s': %s.   Query: %s" % (
                 self.job_name, cname, emsg, sql))
         else: