'All Stats' output contains zeros for all the columns except 'Latency'

[speddam]

Hi,

The latest memtier_benchmark shows the 'All Stats' output containing all zeros for all the columns except 'Latency', like this:

$ ./memtier_benchmark --clients=50 --threads=4 --requests=10000 --data-size-range=32-1000 --pipeline=100 --ratio=5:5
...
ALL STATS
=========================================================================
Type         Ops/sec     Hits/sec   Misses/sec      Latency       KB/sec 
-------------------------------------------------------------------------
Sets            0.00          ---          ---     19.71200         0.00 
Gets            0.00         0.00         0.00     19.73500         0.00 
Waits           0.00          ---          ---      0.00000          --- 
Totals          0.00         0.00         0.00     19.72300         0.00
...

I have done the git bisect and it pointed to this commit [0]:

$ git bisect bad
bec9cc46079476c8414ab7104516c5a9c966b90f is the first bad commit
commit bec9cc46079476c8414ab7104516c5a9c966b90f
Author: YaacovHazan <31382944+YaacovHazan@users.noreply.github.com>
Date:   Wed Dec 11 17:02:15 2019 +0200

    fix issue #94 cluster mode stuck (#106)

    add missing bufferevent_disable for connections that
    complete to send new requests

:100644 100644 b9fe31c3568e172369897c0eea0111138dcc5e8b 6fbcd856130c8709f8786c97e6b38ad101ca342c M	shard_connection.cpp

I'm not familiar with the code base, but by commenting out the bufferevent_disable() call, I see that the output now shows all the columns:

ALL STATS
=========================================================================
Type         Ops/sec     Hits/sec   Misses/sec      Latency       KB/sec 
-------------------------------------------------------------------------
Sets       509744.79          ---          ---     19.52900    281230.68 
Gets       509744.79       509.74    509235.05     19.53200     20145.57 
Waits           0.00          ---          ---      0.00000          --- 
Totals    1019489.58       509.74    509235.05     19.53000    301376.25

---

Regards,
Subhash

[0] bec9cc4?diff=unified#commitcomment-38744660

[rohitjoshi]

I have the same issue and by commenting out the line, it works fine.

@@ -473,7 +473,7 @@ void shard_connection::fill_pipeline(void)
         if (m_conns_manager->hold_pipeline(m_id)) {
             // if we are still connected, disable any event
             if (get_connection_state() == conn_connected) {
-                bufferevent_disable(m_bev, EV_WRITE|EV_READ);
+                //bufferevent_disable(m_bev, EV_WRITE|EV_READ);
             }

             break;

[yaacovhazan-Redislabs]

thank you @speddam and @rohitjoshi for raising the issue and narrow it to the specific line where the issue started... I created PR #130, that should fix it.

[yossigo]

Closing this as I assume it's fixed, please re-open if you still experience a problem.

[godjoem]

Hi, I have a similar issue when running memtier_benchmark with EVAL command in cluster mode. (It's fine when running on a standalone Redis)
The command:

memtier_benchmark -s 127.0.0.1 -p 6380 -c 10 -n 100 --cluster-mode \
  --key-prefix={account}: --key-minimum=1 --key-maximum=10000 \
  --command='EVAL "return redis.call(\"SET\", KEYS[1], ARGV[1])" 1 __key__ __data__' \
  --command-key-pattern=P

Output:

ALL STATS
============================================================================================================================
Type         Ops/sec    MOVED/sec      ASK/sec    Avg. Latency     p50 Latency     p99 Latency    p100 Latency       KB/sec 
----------------------------------------------------------------------------------------------------------------------------
Evals           0.00         0.00         0.00         0.54092         0.39100         3.05500        12.41500         0.00 
Totals          0.00         0.00         0.00         0.54092         0.39100         3.05500        12.41500         0.00 

I use memtier_benchmark 2.1.1 and run Redis cluster in docker containers (latest Redis image).

The output is fine when running the command speddam uses
memtier_benchmark --clients=50 --threads=4 --requests=10000 --data-size-range=32-1000 --pipeline=100 --ratio=5:5 -p 6380 --cluster-mode

I found that this problem occurs when I execute the EVAL command on a Redis cluster.

May you guys kindly help? @YaacovHazan