ignore encoding errors when reading log files

[nominatim.git] / utils / cron_banip.py

diff --git a/utils/cron_banip.py b/utils/cron_banip.py
index d716affbe8e153b99d3fc8810c36ee61dbde9b19..53f5e5f13ab64a1b5527466bc2d84de35b679e73 100755 (executable)
--- a/utils/cron_banip.py
+++ b/utils/cron_banip.py
@@ -52,6 +52,9 @@ LOGFILE= BASEDIR + '/log/restricted_ip.log'
 WHITELIST = ''
 # space-separated list of IPs manually blocked
 BLACKLIST = ''
+# user-agents that should be blocked from bulk mode
+# (matched with startswith)
+UA_BLOCKLIST = ()
 
 # time before a automatically blocked IP is allowed back
 BLOCKCOOLOFF_PERIOD='1 hour'
@@ -61,8 +64,11 @@ BULKCOOLOFF_PERIOD='15 min'
 BULKLONG_LIMIT=8000
 BULKSHORT_LIMIT=2000
 BLOCK_UPPER=19000
-BLOCK_LOADFAC=300
-BULK_LOADFAC=100
+BLOCK_LOWER=4000
+BLOCK_LOADFAC=380
+BULK_LOADFAC=160
+BULK_LOWER=1500
+MAX_BULK_IPS=85
 
 #
 # END OF DEFAULT SETTINGS
@@ -73,11 +79,6 @@ try:
 except IOError:
     pass
 
-# determine current load
-fd = open("/proc/loadavg")
-avgload = int(float(fd.readline().split()[1]))
-fd.close()
-
 # read the previous blocklist
 WHITELIST = set(WHITELIST.split()) if WHITELIST else set()
 prevblocks = []
@@ -99,25 +100,36 @@ try:
 except IOError:
     pass #ignore non-existing file
 
-# current number of bulks
-numbulks = len(prevbulks)
-
-BLOCK_LIMIT = BLOCK_UPPER - BLOCK_LOADFAC * (numbulks - 30)
-BULKLONG_LIMIT = BULKLONG_LIMIT - BULK_LOADFAC * (avgload - 16)
-
+# determine current load
+fd = open("/proc/loadavg")
+avgload = int(float(fd.readline().split()[2]))
+fd.close()
+# DB load
 conn = psycopg2.connect('dbname=nominatim')
 cur = conn.cursor()
+cur.execute("select count(*)/60 from new_query_log where starttime > now() - interval '1min'")
+dbload = int(cur.fetchone()[0])
+
+BLOCK_LIMIT = max(BLOCK_LOWER, BLOCK_UPPER - BLOCK_LOADFAC * (dbload - 75))
+BULKLONG_LIMIT = max(BULK_LOWER, BULKLONG_LIMIT - BULK_LOADFAC * (avgload - 14))
+if len(prevbulks) > MAX_BULK_IPS:
+    BLOCK_LIMIT = max(3600, BLOCK_LOWER - (len(prevbulks) - MAX_BULK_IPS)*10)
+# if the bulk pool is still empty, clients will be faster, avoid having
+# them blocked in this case
+if len(prevbulks) < 10:
+    BLOCK_LIMIT = 2*BLOCK_UPPER
+
 
 # get the new block candidates
 cur.execute("""
-  SELECT ipaddress, max(count) FROM
+  SELECT ipaddress, max(count), max(ua) FROM
    ((SELECT * FROM
-     (SELECT ipaddress, sum(CASE WHEN type = 'search' THEN 3 ELSE 1 END) as count FROM new_query_log
+     (SELECT ipaddress, sum(case when endtime is null then 1 else 1+1.5*date_part('epoch',endtime-starttime) end) as count, substring(max(useragent) from 1 for 30) as ua FROM new_query_log
       WHERE starttime > now() - interval '1 hour' GROUP BY ipaddress) as i
    WHERE count > %s)
    UNION
-   (SELECT ipaddress, count * 4 FROM
-     (SELECT ipaddress, sum(CASE WHEN type = 'search' THEN 2 ELSE 1 END) as count FROM new_query_log 
+   (SELECT ipaddress, count * 3, ua FROM
+     (SELECT ipaddress, sum(case when endtime is null then 1 else 1+1.5*date_part('epoch',endtime-starttime) end) as count, substring(max(useragent) from 1 for 30) as ua FROM new_query_log 
       WHERE starttime > now() - interval '10 min' GROUP BY ipaddress) as i
    WHERE count > %s)) as o
   GROUP BY ipaddress
@@ -125,12 +137,22 @@ cur.execute("""
 
 bulkips = {}
 emergencyblocks = []
+useragentblocks = []
 
 for c in cur:
     if c[0] not in WHITELIST and c[0] not in BLACKLIST:
-        if c[1] > BLOCK_UPPER and c[0] not in prevbulks:
+        # check for user agents that receive an immediate block
+        missing_agent = not c[2]
+        if not missing_agent:
+            for ua in UA_BLOCKLIST:
+                if c[2].startswith(ua):
+                    missing_agent = True
+                    break
+        if (missing_agent or c[1] > BLOCK_UPPER) and c[0] not in prevblocks:
             newblocks.add(c[0])
-            if c[0] not in prevblocks:
+            if missing_agent:
+                useragentblocks.append(c[0])
+            else:
                 emergencyblocks.append(c[0])
         else:
             bulkips[c[0]] = c[1]
@@ -151,15 +173,16 @@ for ip in prevblocks:
         deblockcandidates.add(ip)    
         
 for ip in prevbulks:
-    if ip in bulkips:
-        if bulkips[ip] > BLOCK_LIMIT:
-            newblocks.add(ip)
-            newlyblocked.append(ip)
+    if ip not in newblocks:
+        if ip in bulkips:
+            if bulkips[ip] > BLOCK_LIMIT:
+                newblocks.add(ip)
+                newlyblocked.append(ip)
+            else:
+                newbulks.add(ip)
+            del bulkips[ip]
         else:
-            newbulks.add(ip)
-        del bulkips[ip]
-    else:
-        debulkcandidates.add(ip)
+            debulkcandidates.add(ip)
 
 # cross-check deblock candidates
 if deblockcandidates:
@@ -182,6 +205,7 @@ if debulkcandidates:
     cur.execute("""
         SELECT DISTINCT ipaddress FROM new_query_log
         WHERE ipaddress IN ('%s') AND starttime > now() - interval '%s'
+        AND starttime > date_trunc('day', now())
         """ % ("','".join(debulkcandidates), BULKCOOLOFF_PERIOD))
 
     for c in cur:
@@ -212,6 +236,8 @@ if bulkips:
     fd.write(logstr % ('new bulks:', ', '.join(bulkips.keys())))
 if emergencyblocks:
     fd.write(logstr % ('dir.block:', ', '.join(emergencyblocks)))
+if useragentblocks:
+    fd.write(logstr % (' ua block:', ', '.join(useragentblocks)))
 if newlyblocked:
     fd.write(logstr % ('new block:', ', '.join(newlyblocked)))
 fd.close()