From: Alexandru Avadanii Date: Thu, 24 Aug 2017 03:50:11 +0200 Subject: [PATCH] states: maas: Batch apply linux state with `-b 4` Applying heavy states like linux.network on multiple nodes in parallel might lead to spurious timeouts on slower systems. Compromise between speed and realiability by applying this state in batches of 4 nodes at once. While at it, increase global command_timeout to 15s on Salt master, since default command timeout (10s) is sometimes too small for all minions to return within the required timeslot. Increase it by 50%. Signed-off-by: Alexandru Avadanii --- mcp/config/states/maas | 2 +- mcp/reclass/classes/cluster/baremetal-mcp-ocata-ovs-ha/infra/config.yml | 1 + 2 files changed, 2 insertions(+), 1 deletion(-) diff --git a/mcp/config/states/maas b/mcp/config/states/maas index 12ef9ed..697f83c 100755 --- a/mcp/config/states/maas +++ b/mcp/config/states/maas @@ -107,4 +107,4 @@ wait_for 10 "salt -C '* and not cfg01* and not mas01*' ssh.set_auth_key ${SUDO_U wait_for 10 "salt -C '* and not cfg01* and not mas01*' saltutil.sync_all" wait_for 10 "salt -C '* and not cfg01* and not mas01*' state.apply salt" -wait_for 10 "salt -C '* and not cfg01* and not mas01*' state.apply linux,ntp" +wait_for 10 "salt -C '* and not cfg01* and not mas01*' state.apply linux,ntp -b 4" diff --git a/mcp/reclass/classes/cluster/baremetal-mcp-ocata-ovs-ha/infra/config.yml b/mcp/reclass/classes/cluster/baremetal-mcp-ocata-ovs-ha/infra/config.yml index 7d95ebc..46526da 100644 --- a/mcp/reclass/classes/cluster/baremetal-mcp-ocata-ovs-ha/infra/config.yml +++ b/mcp/reclass/classes/cluster/baremetal-mcp-ocata-ovs-ha/infra/config.yml @@ -37,6 +37,7 @@ parameters: salt: master: accept_policy: open_mode + command_timeout: 15 reclass: storage: data_source: