You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We experienced a Floating point exception in FoundationDB while excluding stateless processes from the cluster.
Cluster details:
FoundationDB 7.1.28
SSD storage engine, double redundancy
36 machines
713 processes in total
10 stateless processes per machine were added in error. We began to remove them per machine, excluding each process one at a time per each machine. At node 27, the floating point exception appeared in fdbcli:
fdb> exclude 10.31.2.88:4579
WARNING: Long delay (Ctrl-C to interrupt)
The database is unavailable; type `status' for more information.
SIGNAL: Floating point exception (8)
Trace: addr2line -e fdbcli.debug -p -C -f -i 0x7ff6d4669980 0xaf2136 0xbcc750 0xbccc60 0xbcd22a 0xbcd486 0xbc9408 0xbc9943 0xbc9d3c 0xbcfda8 0x849a10 0xce1910 0xce1cfb 0x6dda80 0xd91086 0x8adc92 0xc4a5cf 0x869071 0x537923 0x7ff6d4287c87
Floating point exception
The database flipped to unavailable and didn't come back until we re-included the stateless processes and added a process with role = data_distributor.
We experienced a
Floating point exception
in FoundationDB while excluding stateless processes from the cluster.Cluster details:
10 stateless processes per machine were added in error. We began to remove them per machine, excluding each process one at a time per each machine. At node 27, the floating point exception appeared in
fdbcli
:The database flipped to unavailable and didn't come back until we re-included the stateless processes and added a process with
role = data_distributor
.This is a snippet of one of the tracefiles:
I pulled down the debug binaries and ran
addr2line
against them.fdbcli
fdbserver
I see there are OOM's in the tracefiles which I also see in our charts. I'm wondering if this is a known or expected issue?
The text was updated successfully, but these errors were encountered: