diff options
| author |   Zhijiang Hu <hu.zhijiang@zte.com.cn> | 2017-05-04 00:55:33 -0400 |
|---|---|---|
| committer | Zhijiang Hu <hu.zhijiang@zte.com.cn> | 2017-05-04 01:14:50 -0400 |
| commit | 779440ef6064b2d1f2a5004bd92ae3660bc9b63a (patch) | |
| tree | b96a9f13a8d06b700eadb4f1a911699cf15e6588 /ci/build_rpm | |
| parent | 4c57d26e33d6cc4ec66968ab488442eb2e730b45 (diff) | |
Increase PACKETS_PER_BUFFER to 65536
1) This mainly increase PACKETS_PER_BUFFER to 65536 to reduce the
frequency of TCP client acks.
2) Also kills TCP_BUFF_SIZE and define each buffer size in a
more intuitive way.
3) Free more unused memory to prevent being killed by oom-killer
after enlarged PACKETS_PER_BUFFER.
4) Increase client's select() timeout to 20 secs, since we encountered
timeout due to CPU busy in the same BM but with 20 VMs.
Tested this PS in a 10 VM node env, and it can multicast a 2.7G
file to 10 VMs in 6 minutes, while unicast needs 30+ minutes.
Change-Id: Iaf862fb1f1259cc770f720ccdd95dcc281aef262
Signed-off-by: Zhijiang Hu <hu.zhijiang@zte.com.cn>
Diffstat (limited to 'ci/build_rpm')
| -rwxr-xr-x | ci/build_rpm/build_rpms.sh | 4 |
1 files changed, 4 insertions, 0 deletions
diff --git a/ci/build_rpm/build_rpms.sh b/ci/build_rpm/build_rpms.sh index e4b0cae9..2175bd97 100755 --- a/ci/build_rpm/build_rpms.sh +++ b/ci/build_rpm/build_rpms.sh @@ -38,6 +38,10 @@ function cleanup_container { echo "Stopping containers... $containers_to_kill" sudo docker stop -t 2 ${containers_to_kill} echo "Removing containers... $containers_to_kill" + + # Prevent "removal of container daisy is already in progress" + sleep 10 + sudo docker rm -v -f ${containers_to_kill} if [[ ! -z "$volumes_to_remove" ]]; then |