Memory Leak

[Saduff]

Hi,

There seems to be a memory leak. Looking at top, memory usage is steadily increasing, never dropping. Left the computer on over night once, in the morning py3status was using 3.2 GB of memory. 😮
Reloading i3 takes it down to about 30 MB, but in a few hours it will be over 500 MB, always increasing.

Arch Linux 4.12.12-1-ARCH
status_command py3status -c ~/.config/i3/i3status.conf -b -s

Here's my i3status.conf:

general {
    colors = true
}

clock {
    format_time = " %A %d.%m.%Y {icon} %H:%M:%S"
    on_click 1 = "exec gsimplecal"
}

battery_level {
    format = "{icon} {percent}% {time_remaining}"
    notify_low_level = True
    threshold_degraded = 20
}

whatismyip {
    cache_timeout = 1
    negative_cache_timeout = 1
    format = "{icon} {ip}"
}

net_iplist {
    cache_timeout = 1
    iface_blacklist = ["lo", "wlan0"]
    format = "{format_iface}"
    format_iface = "{iface}: {ip4}"
    format_no_ip = "E: down"
    on_click 1 = "exec wicd-client -n"
}

wifi wlan0 {
    cache_timeout = 1
    format_down = " Down"
    format_up = " {bitrate} {signal_percent} {ssid} {ip}"
    on_click 1 = "exec wicd-client -n"
}

diskdata sda1 {
    cache_timeout = 30
    disk = "sda1"
    format = " /{free} GB {used_percent}%"
    on_click 1 = "exec thunar /"
}

sysdata mem {
    cache_timeout = 5
    format = "[\?color=mem  RAM: {mem_used_percent}%]"
}

sysdata cpu {
    cache_timeout = 1
    format = "[\?color=cpu  CPU: {cpu_usage}%]"
}

arch_updates {
    format = "{pacman}"
}

xrandr {
    cache_timeout = 0
    force_on_start = "eDP1=HDMI1"
}

scratchpad_async {
    format = "{counter} ⌫   "
}

order += "scratchpad_async"
order += "arch_updates"
order += "sysdata cpu"
order += "sysdata mem"
order += "diskdata sda1"
order += "wifi wlan0"
order += "net_iplist"
order += "whatismyip"
order += "battery_level"
order += "clock"

[lasers]

Two things to try.

xrandr {
    cache_timeout = 0
    force_on_start = "eDP1=HDMI1"
}

Real-time updates. Every 0 seconds. Use 3600 (hourly) or something else.

scratchpad_async {
    format = "{counter} ⌫   "
}

I personally experienced i3 lagging / freezing with scratchpad_async and/or window_title_async as both modules need to communicate with i3ipc and/or i3. It likely is a bug on their side. Disable this next if you're still experiencing memory leak... as we'd like to rule out xrandr's real-time updates first.

What version? py3status --version. I see we flagged py3status out of date and is still on 3.5 for stable version. Are you using aur/py3status-git? That's all for now.

[Saduff]

I'm using py3status version 3.5 (python 3.6.2) from aur/py3status.

xrandr is not in the order though so it should be disabled? That's a recent change. It was enabled when py3status reached 3.2 GB. But the leak seems to be still there after disabling xrandr earlier today.

I'll try disabling the scratchpad next. What is the normal amount of memory py3status should use?

[lasers]

xrandr is not in the order though so it should be disabled? That's a recent change.

Good catch. Nothing changed. If you're still experiencing the leak after scratchpad, you could try the py3status-git version to catch up on 3.6 and 3.7rc1.

For 3.7, @tobes will be working on "performance consideration and lowering the load footprint of py3status." Your bug reports/performance feedback would help him should you want to participate. :-)

You can try and lighten one more thing.

whatismyip {
    cache_timeout = 1
    negative_cache_timeout = 1
    format = "{icon} {ip}"
}

This would connect to outside https://freegeoip.net/json/ every 1 seconds. Switching to every 60 seconds may help. I think we also have this issue of... For every time we run i3-restart, we get another py3status instance.

[tobes]

Thanks for reporting this, it's something that needs fixing. As @lasers says the nearer we can get to finding this problem the better. If you can find a minimal config that causes the issue that would be really helpful. From my take. The starting point is if this is something core to py3status or a specific module.

[Saduff]

This seems to be in py3status itself. However, modules speed up the leaking significantly.
I left py3status overnight with just the clock module (same config as above). Here are the results:

Time    | Memory usage
--------------------------
start   | 24.9 MB
+ 2h    | 32.7 MB
+ 1h    | 38.8 MB
+ ~12h  | 96.2 MB

And while writing this comment, the memory usage has increased to 97 MB already and increasing slowly as if the garbage collector is on vacation or something.

[tobes]

From what I can see this is a problem with py3status version 3.5 the latest stable version 3.6 does not appear to suffer from the issue I think master is good (this is from very limited testing) we've changed how the modules are run since 3.6

I'm not sure why arch is still on version 3.5 in the aur. I've pinged the arch maintainer @Horgix on irc and here now in case he missed that.

[Saduff]

I installed the git version (3.7rc1) yesterday and left it overnight with just the clock module again.
Seems the problem is still there. Or could it be a regression after 3.6?

Time    | Memory usage
--------------------------
start   | 24.8 MB
+ ~11h  | 70.7 MB

[tobes]

I'm using this to monitor memory usage
while true; do top -bn1 | awk '/py3status/ {print $6}' ; sleep 10; done

I've done these tests using my standard config so they are not terribly scientific. Is it that I have to leave things running much longer to see the issue

3.5 we just keep increasing

28164
28288
28332
28356
28412
29152
29152
29164
29176
29176
29176
29216
30064
30072

master increase but then memory gets released

27528
27532
27532
27544
27544
27548
27576
27576
27576
27580
27584
27584
28288
28324
28324
28324
28324
28324
28324
28336
28336
28276
28276
28240
28240
27516
27516
27516
27516
27400
27400

[Saduff]

I kept 3.7rc1 running for two days with just the clock module. This was the state before day 2:

Time        | Memory usage
--------------------------
midnight    | 111.7 MB (already ran for 1 day)
1 pm (+13h) | 164.6 MB

I ran the same command with my usual config for faster increase. I never see a decrease.

28,1m
28,1m
28,2m
29,1m
29,2m
29,3m
29,5m
29,5m
29,6m
30,5m
30,5m
30,6m
30,8m
30,9m
30,9m
31,2m
31,2m
31,3m
31,6m
31,7m
31,7m
32,0m
32,0m
32,0m
32,9m
33,0m
33,0m
33,3m
33,3m
33,3m
33,5m
33,5m
33,5m
33,9m
34,0m
34,0m
34,3m
34,3m
34,4m
34,5m
34,6m
34,6m
34,9m
34,9m
34,9m
35,4m
35,4m
35,4m
35,7m
35,7m
35,7m
36,0m

[tobes]

ok having run py3status for a long period of time I see memory increases but not to the same level as you are reporting. I need to look at what is the cause. notification limiting is one area where we do cache things permanently which does use memory over time but that should be fairly minimal anyhow a review does no harm.

[Saduff]

It's weird we're getting different results though, as if something external is causing it. I literally left only clock {} in my config and it still goes to 100 MB or above over a longer period of time.

[ultrabug]

Hi guys,

I'd just like to mention that the new release of i3 v4.14.1 fixes a i3bar memory leak.

i3bar: fix a memory leak

@Saduff I dunno which version you're running but the difference may very well be that you're affected by this bug?

Please check it out and report back, thanks!

[Saduff]

My i3 version was 4.14.0. After upgrading to 4.14.1, it seems the leaking has slowed down considerably with my regular py3status config. Still not convinced the problem is gone completely though as I'm still seeing a steady increase, just much slower than before.

---

Yeah, the problem is still there. After running py3status overnight, the memory usage is at 680 MB. With the old i3 version it would have been twice as much though.

[lasers]

Sometimes I leave the computer alone. Sometimes I run i3-msg restart. 200+MB with the latest i3 version. 2 monitors, 4 bars, updated every 900 seconds. http://sprunge.us/YQZX

[mergenelos]

I'm experiencing the same problem on a Lenovo Y510p with i3 version 4.14.1 (2017-09-24) and py3status version 3.7rc1 (python 3.6.2) on a OS: Arch Linux Kernel: x86_64 Linux 4.12.13-1-ARCH.
here is my i3status.conf:

general {
	colors = true
	interval = 10
}


order += "mpd_status"
order += "group netw"
order += "sysdata"
order += "xrandr"
order += "group disks"
order += "bluetooth"
order += "battery_level"
order += "volume_status"
order += "backlight"
order += "keyboard_layout"
order += "dpms"
order += "group today"

group today{
	cycle = 0
	format = " {output}{button}"
	button_next = 1
	fixed_width = True
	tztime local {
		format = "%Y-%m-%d %a %H:%M:%S"
	}
}

dpms{
          icon_off=""
          icon_on=""
}

keyboard_layout{
	format="  {layout}"
	cache_timeout = 0.5
	colors =  "us=#00ffff, ir=#859900"
	on_click 1 ="exec xkblayout-state set +1"
}

backlight{
	format=" {level}%"
	cache_timeout = 1
	button_down = 3
	button_up = 1
	cache_timeout = 0.5
}

volume_status{
        format="  {percentage}%"
        format_muted="  "
        cache_timeout = 0.5
        on_click 1 = "exec pavucontrol"
}

battery_level{
	cache_timeout = 30
	battery_id = "all"
	notify_low_level = True
	threshold_bad = 15
	threshold_degraded = 30
	threshold_full = 100
	format = "{icon} {percent}% {time_remaining}"
	blocks=""
	charging_character=" "
}

bluetooth{
	format=" {name}"
	format_prefix=" "
	format_no_conn="OFF"
	format_no_conn_prefix=" "
	cache_timeout = 1
	color_bad = "#00fff2"
	format = "{name} {mac}"
	on_click 1 = "exec blueman-manager"
}

group disks {
	cycle = 30
	format = "{output} {button}"
	click_mode = "button"
	fixed_width = True
	open = false
	disk "/" {
		format = "  %avail / %total"
	}
	disk "/home" {
		format = "  %avail / %total"
	}
}

xrandr{
	'VGA-1-1_pos' = "left-of eDP-1-1"
	'VGA-1-1_workspaces' = "5: Presentation"
	'HDMI-1-1_pos' = "left-of eDP-1-1"
	'HDMI-1-1_workspaces' = "5: Presentation"
}

sysdata{
	format="[\?color=cpu   {cpu_usage}% {cpu_temp}],[\?color=mem {mem_used}GB ({mem_used_percent}%)]"
	on_click 1 = "exec xfce4-terminal --hide-scrollbar --drop-down --command=htop --title='System Status'"
}

group netw{
	cycle = 0
	format = "{output} {button}"
	fixed_width=true
	open = false
	net_rate{
		format=" {interface} {down} {up}"
		format_no_connection="  "
	}
}

mpd_status{
	cache_timeout=0.5
	max_width=84
	state_pause=""
	state_play=""
	state_stop=""
	#color_pause=""
	#color_play=""
	#color_stop=""
	on_click 1 = "exec --no-startup-id playerctl -p cantata next"
	on_click 2 = "exec --no-startup-id playerctl -p cantata play-pause"
        on_click 3 = "exec --no-startup-id playerctl -p cantata previous"
}

and output of while true; do top -bn1 | awk '/py3status/ {print $6}' ; sleep 10; done

30.4m
33.5m
36.5m
39.4m
42.4m
45.1m
47.7m
50.1m
53.1m
55.9m
58.8m
61.7m
64.5m
67.4m
70.3m
73.2m
76.1m
79.0m
81.9m
84.7m
87.6m
90.4m
93.5m
96.3m
99.2m
102.1m
105.4m
107.7m
110.1m
112.9m
115.8m
118.5m
121.4m
124.3m
127.2m
130.1m
133.0m
135.9m
138.7m
141.6m
144.5m
147.4m
150.3m
153.2m
156.1m
158.9m
161.8m
164.7m
167.5m
170.4m
173.4m
176.3m
179.1m
182.0m
184.9m
187.7m
190.6m
193.4m
196.3m
199.2m
202.2m
205.0m
31.7m //reloaded i3wm

this is 2.9m per 10secs.
I checked memory usage with every single option in my i3status.conf independently. even though memory usage slows down but its still slowly increasing over time.
only time I had no increase (10mins tested) was when i tried:

general {
	colors = true
	interval = 10
}

(removed everything from i3status.conf except lines above).
which only uses 23.1m and stays there...