Measure bloat by continuously updating a table

update-bloat

@@ -0,0 +1,43 @@
+#!/usr/bin/python3
+
+import argparse
+import random
+import time
+
+import psycopg2
+
+ap = argparse.ArgumentParser()
+
+ap.add_argument("--dbname", default="")
+ap.add_argument("--rows", type=int, default=int(1E6))
+ap.add_argument("--updates-per-transaction", type=int, default=1)
+ap.add_argument("--transactions-per-sample", type=int, default=1)
+ap.add_argument("--sleep-between-transactions", type=float, default=1.0)
+args = ap.parse_args()
+
+db = psycopg2.connect("dbname='%s'" % (args.dbname, ))
+csr = db.cursor()
+csr.execute("drop table if exists tmp_update_bloat")
+csr.execute("create table tmp_update_bloat(id serial primary key, t text)")
+for i in range(args.rows):
+    t = "%s.%d" % ("a" * random.randrange(4, 64), i)
+    csr.execute("insert into tmp_update_bloat(t) values(%s)", (t,))
+db.commit()
+
+n_updates = 0
+n_transactions = 0
+while True:
+    csr.execute("select pg_total_relation_size('tmp_update_bloat')")
+    size = csr.fetchone()[0]
+    print(n_transactions, n_updates, size)
+    for i in range(args.transactions_per_sample):
+        for j in range(args.updates_per_transaction):
+            id = random.randrange(args.rows) + 1
+            t = "%s.%d" % ("b" * random.randrange(4, 64), n_updates)
+            csr.execute("update tmp_update_bloat set t = %s where id = %s", (t, id,))
+            n_updates += 1
+        db.commit()
+        n_transactions += 1
+        time.sleep(args.sleep_between_transactions)
+
+