Hi,
We are using Cassandra 3.6 and I have been facing this issue for a while.
When I connect to a cassandra cluster using cqlsh and disconnect the
network keeping cqlsh on, I get really high cpu utilization on client by
cqlsh python process. On network reconnect things return back to normal.
On debugging a particular process with strace I get a lot of lines like:
[pid 8449] connect(4, {sa_family=AF_INET, sin_port=htons(9042),
sin_addr=inet_addr("10.20.34.11")}, 16) = -1 ENETUNREACH (Network is
unreachable)
[pid 8449] close(4) = 0
[pid 8449] futex(0x7f39a8001360, FUTEX_WAKE_PRIVATE, 1) = 1
[pid 5734] <... futex resumed> ) = 0
[pid 5734] futex(0x1956fb0,
FUTEX_WAIT_BITSET_PRIVATE|FUTEX_CLOCK_REALTIME, 0, NULL, ffffffff
<unfinished ...>
[pid 8449] futex(0x1956fb0, FUTEX_WAKE_PRIVATE, 1) = 1
[pid 5734] <... futex resumed> ) = 0
[pid 5734] socket(PF_INET, SOCK_STREAM, IPPROTO_TCP) = 4
[pid 5734] fcntl(4, F_GETFL) = 0x2 (flags O_RDWR)
[pid 5734] fcntl(4, F_SETFL, O_RDWR|O_NONBLOCK) = 0
[pid 5734] connect(4, {sa_family=AF_INET, sin_port=htons(9042),
sin_addr=inet_addr("10.20.34.11")}, 16) = -1 ENETUNREACH (Network is
unreachable)
[pid 5734] close(4) = 0
[pid 5734] socket(PF_INET, SOCK_STREAM, IPPROTO_TCP <unfinished ...>
[pid 8449] futex(0x7f39a40aa390,
FUTEX_WAIT_BITSET_PRIVATE|FUTEX_CLOCK_REALTIME, 0, NULL, ffffffff
<unfinished ...>
Shall I create a jira for the same?
Thanks & Regards,
Bhuvan
