WhatPulse 2.0.2

[dutchmega]

A crash was occuring while pulsing, this has been fixed.
I’ll also be looking into the switching user/remote desktop issue problem.

@Bzzz

Still far too frequent writes to disk!
I have no long-term measurement as 2.02 fails quite often, but 2.01 writes about 80 MB per hour, which is just below the usage of firefox during active browsing. Any other process is negligible against these numbers. 2.02 does about one half to one third of that - great, but not enough. That not only brings noticeable disk load to HDD users, but also has quite a negative effect on SSDs as well. Not only because of the sum of writes, but because it is done very frequently in very small chunks (below page size), so that the real write amount to the NAND is considerably higher, in the region of 1 GB per day on 24/7 systems.

The OS itself should not write it every second with the normal IO/fsync settings. That said, it will be reduced far more in the upcoming versions.

@houkouonchi

I can verify that network usage now appears to be accurate on linux. I When just about saturating a gigabit link I saw 115 megabytes/sec on whatpulse which is what I would expect for saturating a 1 gig link so that is good. Unfortunately it doesn’t look like CPU usage has decreased at all for me (whatpulse still using 30-40% of one core on a core i7 to monitor network).

Hmm I’ll do a few more tests with downloading on Linux. More improvements for this are coming.

[alfio8788]

Today i noticed that sometimes the client stops to count key presses and mouse clicks but doesn’t crash, the only way to make it count again was to close it and start it again.

Noticed this on Windows 8 x64 and XUbuntu 12.04 32bit.

Also noticed that the network stats are still off (at first it seems to count right but then suddently starts falling back), made some testing downloading Debian isos, the first download went right, the second was about 35% off with the generated traffic

[Bloopy]

As of today, 2.0.2 now crashes on startup for me at work (Windows XP), so I’ve had to revert to 2.0.1. Reinstalling 2.0.2 doesn’t make a difference.

This error message shows up when I click debug, if it’s worth anything:

Unhandled exception at 0x67ccfd4e in whatpulse.exe: 0xC0000005: Access violation reading location 0x00000004.

[advcomp2019]

My WhatPulse does not work after playing a game still. I have tried running as admin, and it stops working. If you want to know, I have tested it with Need for Speed World launched from the Origin client. I can test other games from Steam later on.

[afig23]

Whatpulse also seems to randomly crash for me. Sometimes it happens twice in a minute and another time I was running my computer for 3 1/2 days straight and no crash. I do like the reduce HDD usage in this version however.

Win 7 Home Premium 64-Bit
Whatpulse 2.0.2
Here’s one crash:
http://pastebin.com/6NxQn8jB

And here’s another:
http://pastebin.com/XFcy0nm1

[FishGuy876]

Hi,

Testing the new 2.0.2 client here on 3 of my machines, as i’d like to get rid of the HD activity every second. The client crashes on all 3 machines now, seems to happen at night, I have them set to auto pulse around midnight and I think thats when its happening (normally i’m asleep!) I’ve switched back to 2.0.1 as this doesn’t have any crashes at all on my systems.

Keep up the great work guys. The new client is great, and has potential to go a long way once these silly bugs are worked out

Quick Edit: If I try to manually pulse, I also get a crash!

[MetallicA]

I used to have problems resuming after hibernation but with 2.0.2 it seems the problem is fully fixed
kudos and keep the hardwork!

[Sathallrin]

@smitmartijn

• [Linux - Fix] Some network interfaces were not detected if certain system files did not exist

Still nothing for me.

On one machine I get this:

$ ./whatpulse Database check: "ok" Opened database file: "/home/sathallrin/.local/share/data/WhatPulse/whatpulse.db" Creating device listener for "/dev/input/event0" Creating device listener for "/dev/input/event1" Creating device listener for "/dev/input/event10" Creating device listener for "/dev/input/event11" Creating device listener for "/dev/input/event12" Creating device listener for "/dev/input/event13" Creating device listener for "/dev/input/event14" Creating device listener for "/dev/input/event15" Creating device listener for "/dev/input/event16" Creating device listener for "/dev/input/event2" Creating device listener for "/dev/input/event3" Creating device listener for "/dev/input/event4" Creating device listener for "/dev/input/event5" Creating device listener for "/dev/input/event6" Creating device listener for "/dev/input/event7" Creating device listener for "/dev/input/event8" Creating device listener for "/dev/input/event9" Starting NetworkMonitor.. Failed to open: "/sys/class/net/wlan0/device/vendor" or "/sys/class/net/wlan0/device/device" WhatPulse has started

Which is odd since it doesn’t even have a wireless card.

$ ifconfig
eth0: flags=4163<UP,BROADCAST,RUNNING,MULTICAST>  mtu 1500
        inet 10.10.10.131  netmask 255.255.255.0  broadcast 10.10.10.255
        inet6 fe80::21f:c6ff:fe30:41a9  prefixlen 64  scopeid 0x20<link>
        ether 00:1f:c6:30:41:a9  txqueuelen 1000  (Ethernet)
        RX packets 1706117  bytes 2033246229 (1.8 GiB)
        RX errors 0  dropped 0  overruns 0  frame 0
        TX packets 980213  bytes 200388635 (191.1 MiB)
        TX errors 0  dropped 0 overruns 0  carrier 0  collisions 0

lo: flags=73<UP,LOOPBACK,RUNNING>  mtu 65536
        inet 127.0.0.1  netmask 255.0.0.0
        inet6 ::1  prefixlen 128  scopeid 0x10<host>
        loop  txqueuelen 0  (Local Loopback)
        RX packets 316351  bytes 563692968 (537.5 MiB)
        RX errors 0  dropped 0  overruns 0  frame 0
        TX packets 316351  bytes 563692968 (537.5 MiB)
        TX errors 0  dropped 0 overruns 0  carrier 0  collisions 0

On another machine I get this:

$ ./whatpulse 
Database check:  "ok" 
Opened database file:  "/home/sathallrin/.local/share/data/WhatPulse/whatpulse.db" 
Creating device listener for "/dev/input/event0" 
Creating device listener for "/dev/input/event1" 
Creating device listener for "/dev/input/event10" 
Creating device listener for "/dev/input/event2" 
Creating device listener for "/dev/input/event3" 
Creating device listener for "/dev/input/event4" 
Creating device listener for "/dev/input/event5" 
Creating device listener for "/dev/input/event6" 
Creating device listener for "/dev/input/event7" 
Creating device listener for "/dev/input/event8" 
Creating device listener for "/dev/input/event9" 
Starting NetworkMonitor.. 
WhatPulse has started

In both of these cases I get the same thing in 2.0.2 and 2.0.1. They both have eth0 as their network interface.

[borandi]

@Bzzz

Still far too frequent writes to disk!
I have no long-term measurement as 2.02 fails quite often, but 2.01 writes about 80 MB per hour, which is just below the usage of firefox during active browsing. Any other process is negligible against these numbers. 2.02 does about one half to one third of that - great, but not enough. That not only brings noticeable disk load to HDD users, but also has quite a negative effect on SSDs as well. Not only because of the sum of writes, but because it is done very frequently in very small chunks (below page size), so that the real write amount to the NAND is considerably higher, in the region of 1 GB per day on 24/7 systems.

Whatpulse has a very minor effect on SSDs. Reads and Writes can be done on the block level, and are not done on the page level. You are confusing that with deletes, which has to be done on the page level.

We discussed this in depth on this thread: http://whatpulse.org/forums/showthread.php?tid=2008&pid=18208#pid18208

Overall result: Even with an old SSD, it will take several decades of normal WP activity to brick the drive, and only performs about 3% of an ‘low’ daily writes to a drive. Always make sure there is ~20% free on your SSD (general rule) to ensure longevity in all cases.

[commentator]

2.0.2 continues crashing when it tries to autopulse or when you right click on icon and select pulse. Solution is right click, open main window, pres pulse

[Bzzz]

@borandi

Whatpulse has a very minor effect on SSDs. Reads and Writes can be done on the block level, and are not done on the page level. You are confusing that with deletes, which has to be done on the page level.[…]Overall result: Even with an old SSD, it will take several decades of normal WP activity to brick the drive, and only performs about 3% of an ‘low’ daily writes to a drive.[/quote]

No, it doesn’t.
I use full system encryption, so there is a high percentage of used blocks around at all times. Which is generally no problem, as the drive (a 256GB Samsung 830) uses plenty of space for overprovisioning. I bought this drive 19 weeks ago and did some testing until I put it in my machine. These tests did very close to 1 TB of usage. Today, the counter is at 2151 GB. Which equals to ~100 kb/s at 24/7 usage.

As gsmartcontrol offers real-time data from the SSD, I just did two runs of 1 minutes each, one with WP, one without. The diff of written LBAs for the non-WP run was 2144 (~18 kb/s). The other one gave me 14736 LBAs per minute, or ~125 kb/s. That is a HUGE difference. Idle disk load increased sevenfold to 10 gigabytes per day, which is even an order of magnitude larger than I previously expected. Of course you want me to generate statistics over a longer period of time, I’m starting right now…

:edit:
Using this small script →

[code]#!/bin/bash

LBAi=$(sudo smartctl -A /dev/sda | grep 241 | cut -c 88-)
secsi=$(date +"%s")
LBA=$LBAi

echo -e “TIME\t\tLBAs written\tLBA_dT\tLBA_d\tT_dT\tLBAs per minute”
while true
do
LBAl=$LBA
LBA=$(sudo smartctl -A /dev/sda | grep 241 | cut -c 88-)
LBAld=$(expr $LBA - $LBAl)
secs=$(date +"%s")
LBAd=$(expr $LBA - $LBAi)
secsd=$(expr $secs - $secsi)
if (($secsd>0)); then LBAPm=$(expr $LBAd * 60 / $secsd ); fi
echo -e “$secs\t$LBA\t$LBAd\t+$LBAld\t$secsd\t$LBAPm”
sleep 10
done
[/code]
I determined usage patterns for WP:
(Unix time, Total LBAs, diff last LBAs to first ever recorded LBA count, diff LBAs to last sample, time diff in seconds, LBAs per minute since start of the script)

[quote]TIME LBAs written LBA_dT LBA_d T_dT LBAs per minute
1359209676 4638319190 0 +0 0
1359209686 4638322302 3112 +3112 10 18672
1359209787 4638352686 33496 +3096 111 18105
1359209887 4638383854 64664 +3240 211 18387
1359209897 4638386782 67592 +2928 221 18350
1359209988 4638415510 96320 +2848 312 18523
1359210088 4638444630 125440 +2920 412 18267
1359210189 4638473062 153872 +2952 513 17996
1359210289 4638502534 183344 +2872 613 17945[/quote]
and so on, moving to ~17k LBAs per minute (12.5 GiB/day) in long term, with very frequent writes and low deviation.

Now closing WP, I get:

[quote]1359215712 4639489654 0 +0 0
1359215722 4639489830 176 +176 10 1056
1359215923 4639497750 8096 +128 211 2302
1359216124 4639503750 14096 +120 412 2052
1359216325 4639511686 22032 +256 613 2156
1359216526 4639517694 28040 +176 814 2066
1359216727 4639524494 34840 +112 1015 2059
1359216928 4639532454 42800 +184 1216 2111
1359217129 4639538918 49264 +128 1417 2085

Which does have more fluctuations due to other programs writing, but, as you can see, has generally a far lower disk usage than with WP running. 2.1k LBAs per minute is only about 1.5 GiB/day…

[smitmartijn]

Closing thread due to 2.0.3: http://whatpulse.org/forums/showthread.php?tid=2488