In the spirit of this thread, can I ask if there's a way (log somewhere?) to know which layer last printed before a power failure to my octopi, so that I might edit the gcode to start again at following layer? The xy geometry is very small, but the z is very tall so I'd love to complete this print. Also is it safe to assume whatever didn't complete within the last attempted layer is probably ok to skip at 2mm /layer? And finally do I ignore my z offset when calculating new start z-position? Or do I need to include an allowance?
I know of no easy method since OctoPrint pushes many gcode lines into the printer's receive buffer; you don't know how many were actually printed.
Measure accurately, review the gcode file carefully and look at the layer comments. Look at the gcode at the beginning of a layer and look at the G0 or G1 command which includes a Z value.
When I think I've got the layer that's close, I'll then use the manual jogging commands in the Control tab of OctoPrint (no heat) to move the extruder carefully above this point and in some corner. I'll then jog it closer in the Z direction until I've determined what is likely the exact point where it left off.
Taking this back to the gcode, I'll then chop the file's first part to remove it (leaving some of the initial lines). You'd have to turn off autoleveling and any priming lines and any code which would crash into your part. It may be necessary to home the printer and then add a line which brings the extruder assembly somewhere safely over the middle of the part (within your Gcode Start script).
It's best to try to re-do the entire last layer so that there aren't gaps.
It's all very fussy but you can rescue a part.
Great answer - didnt occur to me but its simple and smart - thanks.
I've been accused of being both simple and smart. 
why it say last z heigh so we run repeat it same height 
Unfortunately, OctoPrint sends commands and the firmware & printer board eventually run those commands. So you never know the last command that was run. You're right, though, recording the last z layer/height would help in this.
i say that maybe it say last layer with log.file and i see this panel and i repeat print this layer or last layer or next layer. thiis not importent. because can be sanded
Doesn't the terminal keep track of exactly where you're at in the print? I'm using OctaPrint connected directly to my Ender 3 with no RasPi in between. I've had two failures on the same print today for completely unrelated reasons. The first time we had a general power failure. Time to get a UPS! The second I'm clueless about. The printer said it reset for no good reason while I was sitting at my laptop two feet away.
So my terminal output at the end is:
------------------------
Send: N271889 G0 F300 X111.565 Y115.872 Z140.6*24
Recv: ok
Send: N271890 G0 F9000 X112.292 Y118.917*126
Recv: ok
Send: N271891 G0 X112.695 Y119.788*26
Recv: ok
Send: N271892 G0 X113.275 Y120.554*27
Recv: ok
Send: N271893 G0 X114.009 Y121.176*17
Recv: T:240.00 /240.00 B:59.87 /60.00 @:83 B@:127
Recv: ok
Send: N271894 G0 X114.81 Y121.581*42
Recv: ok
Send: N271895 G0 X116.328 Y122.097*24
Recv: ok
Send: N271896 G1 F1009.3 X116.899 Y122.083 E2789.23934*29
Recv: ok
Send: N271897 G1 X117.047 Y122.062 E2789.24431*101
Recv: ok
Send: N271898 G1 X117.24 Y122.041 E2789.25077*89
Recv: ok
Send: N271899 G1 X117.575 Y121.972 E2789.26214*103
Recv: ok
Send: N271900 G1 X117.732 Y121.932 E2789.26753*101
Recv: start
Printer sent 'start' while printing. External reset? Aborting job since printer lost state.
Send: N0 M110 N0*125
Changing monitoring state from "Printing" to "Cancelling"
Recv: echo: External Reset
Recv: Marlistart
Recv: echo: External Reset
Recv: Marlin Marlin 1.1.6
------------------------
My thought was "Trace back to the last OKed command", which, here, is:
-------------
Send: N271899 G1 X117.575 Y121.972 E2789.26214*103
Recv: ok
-------------
If I find that in my .gcode and search backward for Z, I get:
--------------
Send: N271889 G0 F300 X111.565 Y115.872 Z140.6*24
Recv: ok
--------------
Doesn't that tell me the print stopped at Z=140.6?
I then tried various hacks of the .gcode to remove everything before the last successful command except for heat-up. The first time I deleted everything up to the last successful ("Ok") command then except the preheat commands. However I didn't use G92 to reset the initial position, so I had to flip the power switch before the head went off on some kind of extracurricular adventure.
The second time I added:
G92 X117.575 Y121.972 E2789.26214
...right before the command the print failed on, which is supposed to tell the printer "This is where you current position is."
The second time I added that command but forgot the preheat. First it knocked the print off of the base, but it didn't work anyway and started tracking far away again so I had to shut it off.
I'm not exactly sure why those didn't work, because they weren't just off by a few layers, the head was headed for ZMAX before I hit the power switch, but it sounds like this should work. You leave everything exactly where it is, make sure things don't cool down to where you lose adhesion, and continue the print exactly where it left off. Tomorrow I think I will try this deliberately with a simpler model to see if I can figure out where I went wrong. I also got GetToDaChoppa to try using that to do a re-homed restart, but it seems like doing an in-place restart would be more likely to work.
Love and Spaghetti,
Scott
Restarting a failed job can be tricky. If your printer does an autolevel routine normally, you should turn that off on the 2nd job since otherwise your hotend would slam into your part or get inaccurate readings.
You also want to be careful on parts which sit on a raft. Make sure that you calculate the 2nd job properly.
I wrote GetToDahChoppa to both print multi-color parts by layer and to help resurrect failed parts. It essentially does what I used to do manually: remove some of the lines from the gcode file in the middle section but leaving the start/ending.