Closed totaam closed 3 years ago
Xpra_cmd_2020-03-03_21-19-19.png
(31.4 KiB)xpra-gnome-terminal-latency.png
(16.8 KiB)Xpra-save-graph_cmd_2020-03-03_22-24-26.png
(31.9 KiB)xpra-shadow-high-latency.png
(23.8 KiB)xpra-shadow-high-latency-vivaldi_2020-03-03_21-36-40.png
(35.3 KiB)Please replace the second
It seems though that it's a "generic" issue: [[Image
with this onexpra-gnome-terminal-latency.png
(Another) Side issue:
- Add the filename extension when saving in Windows. Basically, unless the file is saved as
*.png
(in this case), append.png
(even if someone saves with e.g..jpg
)
(idk why fake-xinerama=no, all the others are configured by me server-wide).
Fixed in r25485.
this is an empty gnome terminal, doing nothing. Typing seems like it takes about 3 seconds to manifest.
Please attach xpra info
for the misbehaving session.
UI thread polling waited 27.0 seconds longer than intended (27.5 vs 0.5)
That's not normal at all.
The bajillion "missing icons". Both the logs and the actual missing icons. r25468 does not log anything about missing icons, but they are still missing
That has been fixed in r25463 and tested repeatedly on a blank win10 VM, again today with r25483.
Please replace the second...
I don't understand. You should now have permission to edit the ticket yourself.
Now I noticed it that the functionality only saves one graph (not both of them)
Saving the graph in r25345 froze everything for ~ 30 seconds (^^^^)
r25486 moves the IO to a thread, but the write_to_png
function will still be running in the main thread - and that may be the culprit (PNG is a CPU pig - though 30s is certainly excessive)
Add the filename extension when saving in Windows
Done for all platforms in r25487
Since the server crashed that night (and subsequent debugging tonight did not give more than ~500ms latency), I'll call this my fault.
Apologies.
I'll re-use this for investigating a seamless session: And xpra-info attaching soon
communication-4.png
(16.9 KiB)latency-4.png
(7.2 KiB)redact-xpra-2617-20.log
(442.7 KiB)@Antoine, apologies for "mentioning". I know you are busy; I was wondering if you could pitch in (since I don't know what to look for).
It's annoying that e.g. Teamviewer would work "seamlessly", under any connection. I know I cannot update to v4 (vp9 encoding and stuff), but I guess something could be done 😕
Graph is saying that: [[Image(latency-broken.png)]]
But that's not the image I am getting. Client goes in-and-out of:
2020-04-17 16:53:31,301 server is not responding, drawing spinners over the windows 2020-04-17 16:53:31,556 server is OK again
And this is how server feels like:
>ping 172.16.57.121 Pinging 172.16.57.121 with 32 bytes of data: Reply from 172.16.57.121: bytes=32 time=19ms TTL=61 Reply from 172.16.57.121: bytes=32 time=31ms TTL=61 Reply from 172.16.57.121: bytes=32 time=18ms TTL=61 Reply from 172.16.57.121: bytes=32 time=17ms TTL=61 Ping statistics for 172.16.57.121: Packets: Sent # 4, Received4, Lost = 0 (0% loss), Approximate round trip times in milli-seconds: Minimum # 17ms, Maximum31ms, Average = 21ms
[[Image(netdata_cpu-utilization_vivaldi_2020-04-17_16-54-59.png)]]
(and this was followed by the :20 monitor dying, while :2, :200 are still alive, well and stable)
netdata_cpu-utilization_vivaldi_2020-04-17_16-54-59.png
(50.0 KiB)latency-broken.png
(7.4 KiB)Have you tried setting a
bandwidth-limit
lower than what your ISP gives you? Maybe 10Mbps?
From a recently spawned shadow server:
"Xpra-Python3-x86_64_4.0-26124\xpra_cmd" shadow ssh://user@ip/0 --ssh="plink -ssh -agent" --opengl=no --bandwidth-limit=10Mbps --desktop-scaling=0.75 --min-speed=70 --webcam=no --speaker=off --microphone=off --pulseaudio=no 2020-04-23 10:22:50,479 Xpra GTK3 client version 4.0-26124 64-bit 2020-04-23 10:22:50,482 running on Microsoft Windows 10 2020-04-23 10:22:51,726 GStreamer version 1.16.2 for Python 3.8.2 64-bit 2020-04-23 10:22:52,129 keyboard layout code 0x409 2020-04-23 10:22:52,129 identified as 'United States - English' : us 2020-04-23 10:22:52,629 keyboard settings: layout=us 2020-04-23 10:22:52,635 desktop size is 4160x1440 with 1 screen: 2020-04-23 10:22:52,636 Default (1100x381 mm - DPI: 96x96) workarea: 4160x1400 2020-04-23 10:22:52,636 Generic PnP Monitor 1600x900 at 0x534 (309x174 mm - DPI: 131x131) workarea: 1600x860 2020-04-23 10:22:52,636 C32JG5x 2560x1440 at 1600x0 (697x392 mm - DPI: 93x93) workarea: 2560x1400 2020-04-23 10:22:52,637 downscaled to 75%, virtual screen size: 5547x1920 2020-04-23 10:22:52,637 Default (1100x381 mm - DPI: 128x128) workarea: 5547x1867 2020-04-23 10:22:52,637 Generic PnP Monitor 2133x1200 at 0x712 (309x174 mm - DPI: 175x175) workarea: 2133x1147 2020-04-23 10:22:52,637 C32JG5x 3413x1920 at 2133x0 (697x392 mm - DPI: 124x124) workarea: 3413x1867 2020-04-23 10:23:14,977 enabled remote logging 2020-04-23 10:23:14,979 Xpra GTK3 shadow server version 3.0.9-26132 64-bit 2020-04-23 10:23:14,981 running on Linux Ubuntu 16.04 xenial 2020-04-23 10:23:14,983 remote desktop size is 6400x1440 (xpra_cmd:15532): Pango-WARNING **: 10:23:15.571: couldn't load font "Bitstream Vera Sans Not-Rotated 14.662109375", falling back to "Sans Not-Rotated 14.662109375", expect ugly output. 2020-04-23 10:23:21,598 server is not responding, drawing spinners over the windows 2020-04-23 10:23:23,404 server is OK again 2020-04-23 10:24:53,455 server is not responding, drawing spinners over the windows 2020-04-23 10:24:57,222 server is OK again 2020-04-23 10:24:58,473 server is not responding, drawing spinners over the windows 2020-04-23 10:25:03,498 server is OK again 2020-04-23 10:25:29,915 server is not responding, drawing spinners over the windows 2020-04-23 10:25:37,716 server is OK again 2020-04-23 10:25:43,727 server is not responding, drawing spinners over the windows 2020-04-23 10:25:47,506 server is OK again 2020-04-23 10:25:57,928 server is not responding, drawing spinners over the windows 2020-04-23 10:26:00,705 server is OK again 2020-04-23 10:26:06,139 server is not responding, drawing spinners over the windows 2020-04-23 10:26:17,529 server is OK again 2020-04-23 10:26:23,448 server is not responding, drawing spinners over the windows 2020-04-23 10:26:42,065 server is OK again 2020-04-23 10:26:48,506 server is not responding, drawing spinners over the windows 2020-04-23 10:27:06,663 server is OK again 2020-04-23 10:27:06,833 server is not responding, drawing spinners over the windows 2020-04-23 10:27:07,592 server is OK again 2020-04-23 10:27:13,000 server is not responding, drawing spinners over the windows 2020-04-23 10:27:33,177 server is OK again 2020-04-23 10:27:39,086 server is not responding, drawing spinners over the windows 2020-04-23 10:28:01,005 server is OK again 2020-04-23 10:28:06,173 server is not responding, drawing spinners over the windows 2020-04-23 10:28:18,255 server is OK again 2020-04-23 10:28:24,203 server is not responding, drawing spinners over the windows 2020-04-23 10:28:32,294 server is OK again 2020-04-23 10:28:38,224 server is not responding, drawing spinners over the windows 2020-04-23 10:28:46,342 server is OK again 2020-04-23 10:28:51,507 server is not responding, drawing spinners over the windows 2020-04-23 10:29:22,225 server is OK again 2020-04-23 10:29:27,896 server is not responding, drawing spinners over the windows 2020-04-23 10:29:38,029 server is OK again 2020-04-23 10:30:03,002 server is not responding, drawing spinners over the windows 2020-04-23 10:30:46,846 server is not responding, drawing spinners over the windows 2020-04-23 10:30:47,762 server is OK again 2020-04-23 10:30:47,925 server is OK again 2020-04-23 10:30:53,089 server is not responding, drawing spinners over the windows 2020-04-23 10:31:06,193 server is OK again 2020-04-23 10:31:11,471 server is not responding, drawing spinners over the windows 2020-04-23 10:31:32,258 server is OK again 2020-04-23 10:31:37,200 server is not responding, drawing spinners over the windows 2020-04-23 10:31:47,786 server is OK again 2020-04-23 10:33:44,771 unknown string message: 0xc245 / 0x0 / 0x0 2020-04-23 10:35:36,627 server is not responding, drawing spinners over the windows 2020-04-23 10:35:37,153 server is OK again 2020-04-23 10:35:38,642 server is not responding, drawing spinners over the windows 2020-04-23 10:35:40,153 server is OK again 2020-04-23 10:36:06,150 unknown string message: 0xc245 / 0x0 / 0x0
[[Image(Xpra-2617_cmd_2020-04-23_10-30-18.png)]] [[Image(Xpra-2617_cmd_2020-04-23_10-31-01.png)]]
Xpra-2617_cmd_2020-04-23_10-30-18.png
(32.1 KiB)Xpra-2617_cmd_2020-04-23_10-31-01.png
(26.5 KiB)redact-xpra-2617_2020_04_23_10_32_35.log
(341.3 KiB)From the graphs on comment:11, your server latency spiked to 48 seconds! So something is very very wrong here. It would be interesting to see how much bandwidth xpra was using when things collapsed. From your graph, it was below 16Mbps... This could also be related to the VPN: maybe it's pushing back and not prioritising the xpra traffic enough. My guess is that you should lower the bandwidth limit even more to try to make things work.
I would need to be able to reproduce the problem to have a chance to fix this.
Replying to [comment:12 Antoine Martin]:
From the graphs on comment:11, your server latency spiked to 48 seconds! So something is very very wrong here. It would be interesting to see how much bandwidth xpra was using when things collapsed. From your graph, it was below 16Mbps...
Can you do dual-axis data amount graph? It's not straightforward to translate KB/s to Mbits (and both may be needed)
This could also be related to the VPN: maybe it's pushing back and not prioritising the xpra traffic enough. My guess is that you should lower the bandwidth limit even more to try to make things work.
A crude
iperf
speedtest to my workstation clocks it at ~8 Mbit. I guess I can try to set it to 6 Mbit for the shadow server, but all of that feels wrong 😕I know that I have a v3 server, and extremely big screen to forward. I don't understand why xpra does not / can not lower its quality characteristics to attempt to keep up (server downscaling, color depth etc)
It's true I've set
min-quality=30
,min-speed=50
- I don't know what that means (I've talked about it on a separate ticket I think), but it made some kind of sense visually/responsive-ly at the time. I think that xpra would be able to keep up with quality=30, speed=50.I would need to be able to reproduce the problem to have a chance to fix this.
Can you do dual-axis data amount graph? It's not straightforward to translate KB/s to Mbits (and both may be needed)
Wouldn't it be better to just show the bandwidth in bps? (Mbps / Kbps) And leave the pixels/s alone. Is that what you mean by dual-axis?
I don't understand why xpra does not / can not lower its quality characteristics to attempt to keep up (server downscaling, color depth etc)
It should. I don't understand why it doesn't.
It's true I've set min-quality=30, min-speed=50 - I don't know what that means..
Try min-quality=0
, min-speed=0
.
This will compress harder, at a lower quality, which will use less bandwidth.
Two more things worth trying:
XPRA_SHADOW_REFRESH_DELAY=200 xpra shadow ...
(limits the refresh rate to 5fps - you may even try to increase it even higher)XPRA_MAX_SOFT_EXPIRED=0 xpra shadow ...
(limits how many frames we send without receiving the echo)Important: do this locally, env vars are not forwarded with ssh remote starts.
Replying to [comment:14 Antoine Martin]:
Can you do dual-axis data amount graph? It's not straightforward to translate KB/s to Mbits (and both may be needed) Wouldn't it be better to just show the bandwidth in bps? (Mbps / Kbps) And leave the pixels/s alone. Is that what you mean by dual-axis?
I mean that, on one side you have b/s and on the other bps. I see your point about having only bps - the only annoying thing is when you are trying to translate the units on-the-fly.
Sometimes you are counting size (i.e. bytes), and some times bandwidth (in which "unit-franca" is bps)
I am not saying I have a solid case with b/s, but the fact that conversion is complicated, makes me want both.
I don't understand why xpra does not / can not lower its quality characteristics to attempt to keep up (server downscaling, color depth etc) It should. I don't understand why it doesn't.
It's true I've set min-quality=30, min-speed=50 - I don't know what that means.. Try
min-quality=0
,min-speed=0
. This will compress harder, at a lower quality, which will use less bandwidth.I can try that.
Two more things worth trying:
XPRA_SHADOW_REFRESH_DELAY=200 xpra shadow ...
(limits the refresh rate to 5fps - you may even try to increase it even higher)XPRA_MAX_SOFT_EXPIRED=0 xpra shadow ...
(limits how many frames we send without receiving the echo)Important: do this locally, env vars are not forwarded with ssh remote starts.
Doesn't
--env=
work with those? It is more straightforward than two-step connection.
from xpra info
124 │ 2020-04-29 10:03:30,053 New unix-domain connection received 125 │ 2020-04-29 10:03:30,053 on '/run/xpra/user-precision-t3620-0' 126 │ Traceback (most recent call last): 127 │ File "/usr/lib/python3/dist-packages/xpra/x11/gtk_x11/clipboard.py", line 159, in do_xpra_selection_request 128 │ proxy.do_selection_request_event(event) 129 │ File "/usr/lib/python3/dist-packages/xpra/x11/gtk_x11/clipboard.py", line 359, in do_selection_request_event 130 │ assert requestor 131 │ AssertionError 132 │ 2020-04-29 10:03:30,101 New unix-domain connection received 133 │ 2020-04-29 10:03:30,101 on '/run/user/1000/xpra/user-precision-t3620-0' 134 │ Traceback (most recent call last): 135 │ File "/usr/lib/python3/dist-packages/xpra/x11/gtk_x11/clipboard.py", line 159, in do_xpra_selection_request 136 │ proxy.do_selection_request_event(event) 137 │ File "/usr/lib/python3/dist-packages/xpra/x11/gtk_x11/clipboard.py", line 359, in do_selection_request_event 138 │ assert requestor 139 │ AssertionError 140 │ Traceback (most recent call last): 141 │ File "/usr/lib/python3/dist-packages/xpra/x11/gtk_x11/clipboard.py", line 159, in do_xpra_selection_request 142 │ proxy.do_selection_request_event(event) 143 │ File "/usr/lib/python3/dist-packages/xpra/x11/gtk_x11/clipboard.py", line 359, in do_selection_request_event 144 │ assert requestor 145 │ AssertionError 146 │ Traceback (most recent call last): 147 │ File "/usr/lib/python3/dist-packages/xpra/x11/gtk_x11/clipboard.py", line 159, in do_xpra_selection_request 148 │ proxy.do_selection_request_event(event) 149 │ File "/usr/lib/python3/dist-packages/xpra/x11/gtk_x11/clipboard.py", line 359, in do_selection_request_event 150 │ assert requestor 151 │ AssertionError
Doesn't --env= work with those?
It should. You can verify it did get applied with xpra info
.
assert requestor
That's weird, r26192 converts this to a warning.
Server just dies over and over, it's so frustrating to keep it up
"Xpra-Python3-x86_64_4.0-26124\xpra_cmd" shadow ssh://user@ip/0 --ssh="plink -ssh -agent" --env=XPRA_MAX_SOFT_EXPIRED=0 --opengl=no --bandwidth-limit=6Mbps --desktop-scaling=0.75 --min-speed=70 --webcam=no --speaker=off --microphone=off --pulseaudio=no
and server goes
24 │ 2020-04-29 10:34:58,152 GStreamer version 1.8.3 for Python 3.5.2 64-bit 25 │ 2020-04-29 10:34:58,164 Warning: failed to query pulseaudio using 'pactl info' 26 │ 2020-04-29 10:34:58,164 Connection failure: Connection refused 27 │ 2020-04-29 10:34:58,165 pa_context_connect() failed: Connection refused 28 │ 2020-04-29 10:35:01,184 Warning: cannot load dbus helper: 29 │ 2020-04-29 10:35:01,184 org.freedesktop.DBus.Error.NoServer 30 │ 2020-04-29 10:35:01,184 Failed to connect to socket /tmp/dbus-253Ig9KTwQ 31 │ 2020-04-29 10:35:01,184 Connection refused 32 │ 2020-04-29 10:35:01,192 printer forwarding enabled using pdf and postscript 33 │ 2020-04-29 10:35:01,193 Warning: cannot watch for application menu changes without pyinotify: 34 │ 2020-04-29 10:35:01,193 No module named 'pyinotify' 35 │ 2020-04-29 10:35:01,210 xpra is ready. 36 │ 2020-04-29 10:35:01,210 xpra GTK3 shadow version 3.0.9-26132 64-bit 37 │ 2020-04-29 10:35:01,454 uid=1000 (usr), gid=1000 (usr) 38 │ 2020-04-29 10:35:01,454 running with pid 14800 on Linux Ubuntu 16.04 xenial 39 │ 2020-04-29 10:35:01,455 on display ':0' of size 6400x1440 40 │ 2020-04-29 10:35:01,455 with 3 monitors: 41 │ 2020-04-29 10:35:01,455 DP-0 2560x1440 at 0,0 42 │ 2020-04-29 10:35:01,455 DP-2 1920x1080 at 2560,180 43 │ 2020-04-29 10:35:01,455 DP-4 1920x1080 at 4480,180 44 │ 2020-04-29 10:35:01,540 New unix-domain connection received 45 │ 2020-04-29 10:35:01,541 on '/run/xpra/usr-precision-t3620-0' 46 │ 2020-04-29 10:35:01,668 New unix-domain connection received 47 │ 2020-04-29 10:35:01,679 on '/run/xpra/usr-precision-t3620-0' 48 │ 2020-04-29 10:35:01,739 Handshake complete; enabling connection 49 │ 2020-04-29 10:35:02,231 New unix-domain connection received 50 │ 2020-04-29 10:35:02,232 on '/run/xpra/usr-precision-t3620-0' 51 │ 2020-04-29 10:35:02,237 New unix-domain connection received 52 │ 2020-04-29 10:35:02,237 on '/run/user/1000/xpra/usr-precision-t3620-0' 53 │ 2020-04-29 10:35:02,261 Warning: icon is quite large (273 KB): 54 │ 2020-04-29 10:35:02,261 '/usr/share/pixmaps/com.visualstudio.code.insiders.png' 55 │ 2020-04-29 10:35:03,010 62.7GB of system memory 56 │ 2020-04-29 10:35:04,987 Error setting up client dbus instance: 57 │ 2020-04-29 10:35:04,988 org.freedesktop.DBus.Error.NoServer 58 │ 2020-04-29 10:35:04,988 Failed to connect to socket /tmp/dbus-DTfoSfnVTg 59 │ 2020-04-29 10:35:04,988 Connection refused 60 │ 2020-04-29 10:35:04,989 automatic picture encoding enabled, also available: 61 │ 2020-04-29 10:35:04,989 rgb24, rgb32 62 │ 2020-04-29 10:35:04,990 Python/GTK3 Microsoft Windows 10 aero client version 4.0-26124 64-bit 63 │ 2020-04-29 10:35:04,990 connected 64 │ 2020-04-29 10:35:04,991 Warning: remote clipboard request timed out 65 │ 2020-04-29 10:35:04,991 request id 0, selection=CLIPBOARD, target=TARGETS 66 │ 2020-04-29 10:35:04,996 shadow server: setting default keymap translation 67 │ 2020-04-29 10:35:05,003 client root window size is 5547x1920 68 │ 2020-04-29 10:35:05,053 client 1 @23.198 Xpra GTK3 shadow server version 3.0.9-26132 64-bit 69 │ 2020-04-29 10:35:05,062 client 1 @23.199 running on Linux Ubuntu 16.04 xenial 70 │ 2020-04-29 10:35:05,062 client 1 @23.199 remote desktop size is 6400x1440 71 │ 2020-04-29 10:35:06,610 Warning: not adding duplicate printer 'OneNote' 72 │ 2020-04-29 10:35:06,610 Warning: not adding duplicate printer 'Send To OneNote 16' 73 │ 2020-04-29 10:35:06,611 Warning: not adding duplicate printer 'Microsoft Print to PDF' 74 │ 2020-04-29 10:35:45,112 Warning: remote clipboard request timed out 75 │ 2020-04-29 10:35:45,113 request id 2, selection=CLIPBOARD, target=UTF8_STRING 76 │ 2020-04-29 10:36:55,764 client 1 @30.081 server is not responding, drawing spinners over the windows 77 │ 2020-04-29 10:36:55,769 client 1 @03.359 server is OK again 78 │ 2020-04-29 10:36:55,769 client 1 @05.104 server is not responding, drawing spinners over the windows 79 │ 2020-04-29 10:36:55,770 client 1 @05.358 server is OK again 80 │ 2020-04-29 10:36:55,770 client 1 @13.241 server is not responding, drawing spinners over the windows 81 │ 2020-04-29 10:36:55,819 Warning: timeout on screen updates for window 2, 82 │ 2020-04-29 10:36:55,819 already delayed for more than 108 seconds 83 │ 2020-04-29 10:36:57,312 Warning: request id 2 not found 84 │ 2020-04-29 10:36:57,312 timed out already? 85 │ 2020-04-29 10:36:57,318 xpra client 1 disconnected. 86 │ 2020-04-29 10:37:36,709 ignoring token for disabled clipboard 'CLIPBOARD' 87 │ 2020-04-29 10:37:36,709 ignoring token for disabled clipboard 'CLIPBOARD' 88 │ 2020-04-29 10:37:36,720 Last client has disconnected, terminating 89 │ 2020-04-29 10:37:36,879 removing socket '/run/user/1000/xpra/usr-precision-t3620-0' 90 │ 2020-04-29 10:37:36,879 removing socket '/run/xpra/usr-precision-t3620-0' 91 │ xpra initialization error: 92 │ cannot find live server for display :0 93 │ xpra initialization error: 94 │ cannot find live server for display :0
It seems that only
XPRA_SHADOW_REFRESH_DELAY
does the thing for me - with 2 out of 3 monitors/windows minimized at all times. Which is working really superb-lyHowever, I have so many objections to what's happening.
This cannot be "very low quality": [[Image(xpra-shadow_Xpra_cmd_2020-04-29_10-43-22.png)]] [[Image(xpra-2617-all-shadow.png)]]
and definitely not 5fps (It would be too much to shoot a video, I guess you can take my word for it). From gaming experience, that would be a table-flip, not a seamless experience (disregarding the 1/3 monitors available issue).
Even though it seems applied:
$ xpra info :0 | grep XPRA_SHADOW_REFRESH_DELAY env.XPRA_SHADOW_REFRESH_DELAY=200
However, I cannot find it by "conventional" means:
$ ps faux | grep '/usr/bin/python3 /usr/bin/xpra' | grep :0 sntentos 3027 0.0 0.0 276660 23188 ? Sl 09:50 0:01 | \_ /usr/bin/python3 /usr/bin/xpra _proxy :2 sntentos 22382 0.3 0.0 276664 23492 ? Sl 10:41 0:01 \_ /usr/bin/python3 /usr/bin/xpra _proxy_shadow_start :0 --opengl=no --bandwidth-limit=6Mbps --ssh=plink -ssh -agent --desktop-scaling=0.75 --speaker=off --webcam=no --min-speed=70 --env=XPRA_SHADOW_REFRESH_DELAY=200 sntentos 3487 0.0 0.0 15456 984 pts/25 S+ 10:51 0:00 | \_ grep --color=auto /usr/bin/python3 /usr/bin/xpra sntentos 4859 0.0 0.0 276664 16900 ? Sl Απρ14 0:04 \_ /usr/bin/python3 /usr/bin/xpra _proxy :2 sntentos 7754 0.0 0.0 276664 16964 ? Sl Απρ16 0:04 \_ /usr/bin/python3 /usr/bin/xpra _proxy :2 sntentos 25943 0.0 0.0 276664 18480 ? Sl Απρ20 0:00 \_ /usr/bin/python3 /usr/bin/xpra _proxy :2 sntentos 26512 0.0 0.0 276664 18568 ? Sl Απρ22 0:02 \_ /usr/bin/python3 /usr/bin/xpra _proxy :200 sntentos 14542 0.0 0.0 276676 18356 ? Sl Απρ22 0:07 \_ /usr/bin/python3 /usr/bin/xpra _proxy :2 root 22101 0.0 0.0 325340 27644 ? Ssl Απρ23 0:01 /usr/bin/python3 /usr/bin/xpra proxy :14500 --daemon=no --bind-tcp=0.0.0.0:14500 --tcp-auth=sys --ssl-cert=/etc/xpra/ssl-cert.pem --ssl=on --bind=/run/xpra/system --auth=peercred --socket-dirs=/run/xpra --socket-permissions=666 --log-dir=/var/log --pidfile=/run/xpra.pid --debug= sntentos 31999 0.0 0.0 276672 23492 ? Sl Απρ28 0:02 \_ /usr/bin/python3 /usr/bin/xpra _proxy :20 sntentos 22414 68.6 0.7 2068812 462560 ? Sl 10:41 6:11 /usr/bin/python3 /usr/bin/xpra shadow :0 --speaker=off --desktop-scaling=0.75 --bandwidth-limit=6Mbps --exit-with-client=yes --min-speed=70 --opengl=no --ssh=plink -ssh -agent --env=XPRA_SHADOW_REFRESH_DELAY=200 --env=XPRA_PROXY_START_UUID=21a59d1f24eb42a382263c574ea7dc60 --daemon=yes --systemd-run=no $ PID=22414 $ cat "/proc/${PID}/environ" | tr '\0' '\n' | grep -i XPRA_SHADOW_REFRESH_DELAY $
xpra-2617-all-shadow.png
(330.6 KiB)xpra-shadow_Xpra_cmd_2020-04-29_10-43-22.png
(1866.8 KiB)Which is working really superb-ly (..) This cannot be "very low quality":
You are setting the minimum quality, auto-refresh will still sharpen the image when it is given time to do so. (ie: when the picture is not changing much - like the login screen)
and definitely not 5fps (It would be too much to shoot a video, I guess you can take my word for it). From gaming experience, that would be a table-flip, not a seamless experience
I thought it was working superbly? What fps are you getting? Shadow is not targeting a gaming-like experience for sure, especially not with 6Mbps and large screens.
However, I cannot find it by "conventional" means:
It is in both shadow processes (shadow
and the remote _proxy_shadow_start
).
Replying to [comment:20 Antoine Martin]:
Which is working really superb-ly (..) This cannot be "very low quality": You are setting the minimum quality, auto-refresh will still sharpen the image when it is given time to do so. (ie: when the picture is not changing much - like the login screen)
The only qualities I saw during testing was the fine-crisp login screen / terminal images above, and the spinner. (the other app was Evolution, and the WQHD monitor was showing the empty desktop)
and definitely not 5fps (It would be too much to shoot a video, I guess you can take my word for it). From gaming experience, that would be a table-flip, not a seamless experience I thought it was working superbly? What fps are you getting? Shadow is not targeting a gaming-like experience for sure, especially not with 6Mbps and large screens.
Oh yeah - that one was misleading, apologies. It is working superbly. While there was no fps counter, I feel fairly confident to say it felt like a solid 20fps.
However, I cannot find it by "conventional" means: It is in both shadow processes (
shadow
and the remote_proxy_shadow_start
).
I also don't understand - why if the bandwidth limitation is a problem (i.e. I am maxing out the line and "server is not responding" because of no bandwidth left), why isn't the bandwidth maxed out as a flat line to a point, and instead it's a 0-flat line?
[[Image(spped-1.png)]]
... Actually, that is not the I image I clicked to save. I'll take only screenshots instead.
spped-1.png
(17.7 KiB)Server just dies over and over, it's so frustrating to keep it up
Could it be #2730?
Replying to [comment:23 Antoine Martin]:
Server just dies over and over, it's so frustrating to keep it up Could it be #2730?
No, it's the by-default exit-with-client 😕
and definitely not 5fps
Are you sure that you applied the environment variables separately and correctly?
With XPRA_SHADOW_REFRESH_DELAY=200
, I am unable to get more than 5fps and I really don't see how it would be possible to do so.
What is there more "true" than your own diagnostics? 😕
$ xpra info :0 | grep XPRA_SHADOW_REFRESH_DELAY env.XPRA_SHADOW_REFRESH_DELAY=200
I can try to check it again. And I'll add
--exit-with-client=no
for good measure
I can try to check it again.
Please include the full xpra info
.
Apologies for the tarball, but I did not have the patience to upload 8 different items with this uploader 😕
So, a quick recap, in archive's order:
- NOT 5fps video (unless I cannot see it, idk)
- Too high client CPU for 5fps (or in general looks too high, but I think we've discussed you cannot fix this against Xenial Server)
- Graph says recv 32000 Kb/s. (a) I doubt it, (b) I have 6mbps limit set
- Graph shows how "server is not responding, drawing spinners over the windows" looks like to me - 0 received data. How does that even work?
- Latency seems off. On "Server not responding", can it just render blank? I would say 9999, but then that will overshoot y-axis until stabilization
- Client and redacted xpra info (collected at Xpra-2617-weird-graph_cmd_2020-05-15_11-16-51.png's time)
xpra-2617-diag-v2.zip
(196.4 KiB)Another instance of a seamless server excessively lagging. Sadly, I lost the client (and the log 😕)
[[Image(xpra-lag-200.png)]]
Example would look like this:
"Xpra-Python3-x86_64_4.0-26306\xpra_cmd" attach ssh://user@ip/200 --ssh="plink -ssh -agent" --modal-windows=no --title="@title@ on @@/@server-display@" --opengl=no --bandwidth-limit=6Mbps 2020-06-04 12:41:12,078 Xpra GTK3 client version 4.0-26306 64-bit 2020-06-04 12:41:12,080 running on Microsoft Windows 10 2020-06-04 12:41:12,170 Warning: failed to import opencv: 2020-06-04 12:41:12,171 No module named 'cv2' 2020-06-04 12:41:12,171 webcam forwarding is disabled 2020-06-04 12:41:13,037 GStreamer version 1.16.2 for Python 3.8.2 64-bit 2020-06-04 12:41:13,582 keyboard layout code 0x409 2020-06-04 12:41:13,583 identified as 'United States - English' : us 2020-06-04 12:41:13,939 keyboard settings: layout=us 2020-06-04 12:41:13,944 desktop size is 4160x1440 with 1 screen: 2020-06-04 12:41:13,945 Default (1100x381 mm - DPI: 96x96) workarea: 4160x1400 2020-06-04 12:41:13,946 Generic PnP Monitor 1600x900 at 0x534 (309x174 mm - DPI: 131x131) workarea: 1600x860 2020-06-04 12:41:13,946 C32JG5x 2560x1440 at 1600x0 (697x392 mm - DPI: 93x93) workarea: 2560x1400 2020-06-04 12:41:35,150 enabled remote logging 2020-06-04 12:41:35,154 Xpra GTK3 X11 server version 3.0.9-26132 64-bit 2020-06-04 12:41:35,156 running on Linux Ubuntu 16.04 xenial 2020-06-04 12:41:35,187 Attached to ip:22 2020-06-04 12:41:35,188 (press Control-C to detach) (xpra_cmd:18576): Pango-WARNING **: 12:41:35.866: couldn't load font "Bitstream Vera Sans Not-Rotated 14.662109375", falling back to "Sans Not-Rotated 14.662109375", expect ugly output. 2020-06-04 12:41:56,477 server is not responding, drawing spinners over the windows 2020-06-04 12:42:00,244 server is OK again 2020-06-04 12:42:05,087 toggling debug on backing gtk3.CairoBacking(<cairo.ImageSurface object at 0x0000000003fe3c50>) for window 286 2020-06-04 12:42:12,778 server is not responding, drawing spinners over the windows 2020-06-04 12:42:17,293 server is OK again 2020-06-04 12:42:53,314 server is not responding, drawing spinners over the windows 2020-06-04 12:43:03,372 server is OK again 2020-06-04 12:44:17,635 unknown string message: 0xc2a9 / 0x0 / 0x0 2020-06-04 12:44:43,498 server is not responding, drawing spinners over the windows 2020-06-04 12:44:51,293 server is OK again 2020-06-04 12:44:59,257 server is not responding, drawing spinners over the windows 2020-06-04 12:45:08,649 unknown string message: 0xc2a9 / 0x0 / 0x0 2020-06-04 12:45:11,012 server is OK again 2020-06-04 12:45:59,852 server is not responding, drawing spinners over the windows 2020-06-04 12:46:14,220 server is OK again 2020-06-04 12:46:47,026 unknown string message: 0xc2a9 / 0x0 / 0x0 2020-06-04 12:49:09,459 server is not responding, drawing spinners over the windows 2020-06-04 12:49:14,252 server is OK again 2020-06-04 12:49:19,664 server is not responding, drawing spinners over the windows 2020-06-04 12:49:51,794 server is OK again 2020-06-04 12:50:16,892 unknown string message: 0xc2a9 / 0x0 / 0x0
... followed by "server connection lost", with version varying in the r26160 - r26503 range
redact-display-_200-$TIMESTAMP.log.bz2
(378.6 KiB)redact-xpra-lag-200.log
(169.3 KiB)xpra-lag-200.png
(38.3 KiB)Another instance of a seamless server excessively lagging. Sadly, I lost the client (and the log 😕) This ticket is about the
shadow
server, is this now happening with theseamless
server too? Can you grab thexpra info
when the client is connected and having problems? 2.5 seconds between each screen update is worse than slideshow!
Replying to [comment:31 Antoine Martin]:
Another instance of a seamless server excessively lagging. Sadly, I lost the client (and the log 😕) This ticket is about the
shadow
server, is this now happening with theseamless
server too?I started mixing them, given that I see a lot less of shadow server action (I gave up trying to make that work as my default), and that issue appears in seamless modes too.
What is more annoying, is that e.g., for every 5 invocations, 2 are broken, 1 is meh, and 1 or 2 are maybe working (instead e.g. for something clearly, constantly broken). In seamless modes, those limitations are changed to 1 broken link every 15 invocations.
Can you grab the
xpra info
when the client is connected and having problems?Apologies,
redact-xpra-lag-200.log
is an xpra-info. Added bonus: the timeout messages given when I was collectingxpra info
concurrently with the client trying not to choke.2.5 seconds between each screen update is worse than slideshow!
If you'd plot server timeout values like I asked somewhere (using a magic 9999 number), given wide enough graphs, that graph would look like a cardiograph instead. "Slideshow" is a gross understatement.
Issue migrated from trac ticket # 2617
component: client | priority: major
2020-03-03 20:31:54: stdedos created the issue