-
Notifications
You must be signed in to change notification settings - Fork 365
Open
Description
The metrics logging is a bit out of control and excessively logs.
# grep solana_metrics::metrics firedancer-since-09:30.log | grep -c 09:30:01
1328With an agave instance, I can do RUST_LOG=info,solana_metrics=warn to suppress this, but I am unable to do the same for firedancer as the sandbox seems to be clearing them:
firedancer/src/util/sandbox/fd_sandbox.c
Lines 122 to 132 in 076b3cf
| void FD_FN_SENSITIVE | |
| fd_sandbox_private_explicit_clear_environment_variables( void ) { | |
| if( !environ ) return; | |
| for( char * const * env = environ; *env; env++ ) { | |
| ulong len = strlen( *env ); | |
| explicit_bzero( *env, len ); | |
| } | |
| if( clearenv() ) FD_LOG_ERR(( "clearenv failed" )); | |
| } |
Can we add support for setting the env variables RUST_LOG and RUST_BACKTRACE to frankendancer for the agave bits until the full client is live?
I set that variable in the firedancer systemd unit and it didn't seem to be working. I confirmed this behaviour via:
# cat /proc/$(pgrep -f 'fdctl.*run-agave')/environ
#This is an example of a single one of the log lines:
firedancer[20839]: INFO 2025-11-29 09:30:00.015801548 GMT+00 20838:7 firedancer:[${HOSTNAME_HERE}] :f10 fd1:[group]:0 metrics/src/metrics.rs(331)[solana_metrics::metrics]: datapoint: slot_stats_tracking_complete slot=128572897i last_index=63i num_repaired=0i num_recovered=0i min_turbine_fec_set_count=64i is_full=true is_rooted=true is_dead=false
Metadata
Metadata
Assignees
Labels
No labels