Setting up another guppy server

[looselab]

Setting up another guppy server

[alexomics]

@tchrisboles If I understand you correctly, yes.

You can setup another guppy_basecall_server on another port (we generally choose 5556) on the same computer and point ru at it. For example, in the TOML the caller_settings table will look like:

[caller_settings]
config_name = "dna_r9.4.1_450bps_fast"
host = "127.0.0.1"
port = 5556

To set up the server I would start a tmux or screen session and run something like so:

$ /path/to/guppy_basecall_server --config dna_r9.4.1_450bps_fast.cfg \
    --port 5556 --log_path . --ipc_threads 1 --max_queued_reads 2000 \
    --chunks_per_runner 48 --num_callers 1 -x cuda:all --gpu_runners_per_device 2

[tchrisboles]

Great - will try it today and tomorrow and report. Thanks

[tchrisboles]

Alex and Matt, I have made some progress on setting up my machine. Minknow seems to run your simulation fast5 bulk file. The unblock all test seems to work as your readme shows. However the enrich chr21,22 doesn't seem to work. The ru_summarise_fq report doesn't show the expected enhancement of mean read lengths for both chromosomes. Chr21 is slighly enhanced (but not as much as your example), and chr22 is not enhanced at all.

After 22 min of ru running using the simulation the size distributions look like:


The toml file was modified setting max_chunks to 12

I set up the guppy basecall server in a dedicated terminal window before starting Minknow like this:

Then started Minknow simulation run using your bulk fast5 file. Once the simulation started I opened a new terminal window and started ru_generators:

After 15 minutes, the mapping speeds were fairly constant, ranging between 0.08 and 0.15s.

My gpu is an RTX2080ti with 11Gb memory.
The workstation is an HP-Z8-G4 with dual Xeon processors (each 10 core), 128Gb ram, 1Tb ssd.

[tchrisboles]

Changing to guppy server 3.4.5 seems to solve this issue - see my post in question #44.

[tchrisboles]

Alex - in view of progress on question #44 - I think this one is closed.