)]}'
{"/COMMIT_MSG":[{"author":{"_account_id":3217,"name":"Ian Wells","username":"ijw-ubuntu"},"change_message_id":"957866199b8f8b7e074af23cecb4d9f816d0f104","unresolved":false,"context_lines":[{"line_number":6,"context_line":""},{"line_number":7,"context_line":"Configure mtu for ovs with the common protocols"},{"line_number":8,"context_line":""},{"line_number":9,"context_line":"When you are using (datagram) tunneling the recommended/working guest device"},{"line_number":10,"context_line":"mtu is less than your tunneling device mtu by the tunneling protocols"},{"line_number":11,"context_line":"header sizes. This change calculates the well working mtu size for gre"},{"line_number":12,"context_line":"and vxlan tunneling and allow you to specify Q_GUEST_DEVICE_MTU for"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":6,"id":"9a80dd14_f3ee44ec","line":9,"updated":"2015-03-06 22:16:19.000000000","message":"L3 tunneling, specifically - VLANs are not addressed by this patch, and generally don\u0027t change the MTU anyway (and honestly the way the patch is written it\u0027s not clear that was a conscious connection you made).","commit_id":"d1534734b6e6b9d492efaa1e52a375938ea53184"},{"author":{"_account_id":3217,"name":"Ian Wells","username":"ijw-ubuntu"},"change_message_id":"957866199b8f8b7e074af23cecb4d9f816d0f104","unresolved":false,"context_lines":[{"line_number":16,"context_line":"devstack will not try to configure mtu settings."},{"line_number":17,"context_line":""},{"line_number":18,"context_line":"When the \u0027openvswitch\u0027 agent binary used together with ovs, the mtu size"},{"line_number":19,"context_line":"advertised via dhcp also configured."},{"line_number":20,"context_line":"Probably this would be doable with other ovs user agents."},{"line_number":21,"context_line":""},{"line_number":22,"context_line":"Change-Id: Idc97dee0aac5792b96994e5f7b524deafbb20a58"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":6,"id":"9a80dd14_93b7c8f4","line":19,"updated":"2015-03-06 22:16:19.000000000","message":"Why wouldn\u0027t this work with a linuxbridge driver? The DHCP agent is independent.","commit_id":"d1534734b6e6b9d492efaa1e52a375938ea53184"},{"author":{"_account_id":5803,"name":"Attila Fazekas","email":"afazekas@redhat.com","username":"afazekas"},"change_message_id":"c4f42b8ffaa00493bca00af1138e7276b6d85e3e","unresolved":false,"context_lines":[{"line_number":16,"context_line":"devstack will not try to configure mtu settings."},{"line_number":17,"context_line":""},{"line_number":18,"context_line":"When the \u0027openvswitch\u0027 agent binary used together with ovs, the mtu size"},{"line_number":19,"context_line":"advertised via dhcp also configured."},{"line_number":20,"context_line":"Probably this would be doable with other ovs user agents."},{"line_number":21,"context_line":""},{"line_number":22,"context_line":"Change-Id: Idc97dee0aac5792b96994e5f7b524deafbb20a58"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":6,"id":"9a80dd14_bf0ac623","line":19,"in_reply_to":"9a80dd14_93b7c8f4","updated":"2015-03-07 06:38:59.000000000","message":"There is quadrillion of neutron driver, the ml2-ovs-vxlan is the most commonly used in the dev envs including the gate jobs, so this is the primary target for now.","commit_id":"d1534734b6e6b9d492efaa1e52a375938ea53184"},{"author":{"_account_id":3217,"name":"Ian Wells","username":"ijw-ubuntu"},"change_message_id":"f0aec1c81fe036b741e7f8e6a72c837b659419c3","unresolved":false,"context_lines":[{"line_number":16,"context_line":"devstack will not try to configure mtu settings."},{"line_number":17,"context_line":""},{"line_number":18,"context_line":"When the \u0027openvswitch\u0027 agent binary used together with ovs, the mtu size"},{"line_number":19,"context_line":"advertised via dhcp also configured."},{"line_number":20,"context_line":"Probably this would be doable with other ovs user agents."},{"line_number":21,"context_line":""},{"line_number":22,"context_line":"Change-Id: Idc97dee0aac5792b96994e5f7b524deafbb20a58"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":6,"id":"9a80dd14_451bb872","line":19,"in_reply_to":"9a80dd14_bf0ac623","updated":"2015-03-07 23:02:45.000000000","message":"How would a user of devstack know that it works for some configurations of OVS, and not other devstack-supported network configs?  It\u0027s not brought to their attention.","commit_id":"d1534734b6e6b9d492efaa1e52a375938ea53184"}],"functions-common":[{"author":{"_account_id":970,"name":"Dean Troyer","email":"dtroyer@gmail.com","username":"dtroyer"},"change_message_id":"1e071ba4faa592b242b41934a799ed1fd55cb377","unresolved":false,"context_lines":[{"line_number":661,"context_line":"# returns empty string when not found"},{"line_number":662,"context_line":"function get_if_for_ip {"},{"line_number":663,"context_line":"    local ip\u003d$1"},{"line_number":664,"context_line":"    ip -o address | grep $ip | awk \u0027{print $2}\u0027"},{"line_number":665,"context_line":"}"},{"line_number":666,"context_line":""},{"line_number":667,"context_line":"# get the mtu of network interface"}],"source_content_type":"application/octet-stream","patch_set":1,"id":"baada198_f5402161","line":664,"updated":"2014-08-08 04:19:52.000000000","message":"Go ahead and just use awk for the grep function too.","commit_id":"3340fd4af2bc4b325eaa8c40a4d203e83325bc0e"},{"author":{"_account_id":970,"name":"Dean Troyer","email":"dtroyer@gmail.com","username":"dtroyer"},"change_message_id":"1e071ba4faa592b242b41934a799ed1fd55cb377","unresolved":false,"context_lines":[{"line_number":665,"context_line":"}"},{"line_number":666,"context_line":""},{"line_number":667,"context_line":"# get the mtu of network interface"},{"line_number":668,"context_line":"function get_mac_for_if {"},{"line_number":669,"context_line":"    local iface\u003d$1"},{"line_number":670,"context_line":"    ip -o link show \"$iface\" | sed \u0027s/.*mtu \\([[:digit:]][[:digit:]]*\\) .*/\\1/\u0027"},{"line_number":671,"context_line":"}"}],"source_content_type":"application/octet-stream","patch_set":1,"id":"baada198_55d2adcf","line":668,"updated":"2014-08-08 04:19:52.000000000","message":"Should this be named get_mtu_for_if?","commit_id":"3340fd4af2bc4b325eaa8c40a4d203e83325bc0e"},{"author":{"_account_id":3217,"name":"Ian Wells","username":"ijw-ubuntu"},"change_message_id":"957866199b8f8b7e074af23cecb4d9f816d0f104","unresolved":false,"context_lines":[{"line_number":744,"context_line":"}"},{"line_number":745,"context_line":""},{"line_number":746,"context_line":"# get the mtu size for an ip"},{"line_number":747,"context_line":"# The ip address has to be associated to a network interface on the current host"},{"line_number":748,"context_line":"function get_mtu_for_ip {"},{"line_number":749,"context_line":"    local ip\u003d$1"},{"line_number":750,"context_line":"    local iface\u003d$(get_if_for_ip \"$ip\")"}],"source_content_type":"application/x-shellscript","patch_set":6,"id":"9a80dd14_1346784f","line":747,"updated":"2015-03-06 22:16:19.000000000","message":"This is the segment MTU size for the segment that the interface is on, assuming I can find an interface with this source IP and also making a few key assumptions about how the network interfaces are configured.  (So in fact it\u0027s probably the segment MTU but it could certainly be wrong.)  But it is not the path MTU, which is what actually affects L3 encapsulations.\n\nPlaying the odds, it\u0027s *probably* going to work, but it\u0027s certainly not guaranteed to be right and it\u0027s actually very hard to prove it\u0027s wrong - which means even with this change devstack will mess up the MTU of deployments in a way the user will discover at a late stage.","commit_id":"d1534734b6e6b9d492efaa1e52a375938ea53184"},{"author":{"_account_id":5803,"name":"Attila Fazekas","email":"afazekas@redhat.com","username":"afazekas"},"change_message_id":"c4f42b8ffaa00493bca00af1138e7276b6d85e3e","unresolved":false,"context_lines":[{"line_number":744,"context_line":"}"},{"line_number":745,"context_line":""},{"line_number":746,"context_line":"# get the mtu size for an ip"},{"line_number":747,"context_line":"# The ip address has to be associated to a network interface on the current host"},{"line_number":748,"context_line":"function get_mtu_for_ip {"},{"line_number":749,"context_line":"    local ip\u003d$1"},{"line_number":750,"context_line":"    local iface\u003d$(get_if_for_ip \"$ip\")"}],"source_content_type":"application/x-shellscript","patch_set":6,"id":"9a80dd14_32060d1e","line":747,"in_reply_to":"9a80dd14_1346784f","updated":"2015-03-07 06:38:59.000000000","message":"The patch main goal is to help in very simple 2-3 vm or 2-3 physical machine setups, including multinod- gate jobs, with the most commonly used dev setup.\n\nIn most cases these (host) test machines gets the right MTU from their dhcp server (for example from openstack).","commit_id":"d1534734b6e6b9d492efaa1e52a375938ea53184"},{"author":{"_account_id":3217,"name":"Ian Wells","username":"ijw-ubuntu"},"change_message_id":"f0aec1c81fe036b741e7f8e6a72c837b659419c3","unresolved":false,"context_lines":[{"line_number":744,"context_line":"}"},{"line_number":745,"context_line":""},{"line_number":746,"context_line":"# get the mtu size for an ip"},{"line_number":747,"context_line":"# The ip address has to be associated to a network interface on the current host"},{"line_number":748,"context_line":"function get_mtu_for_ip {"},{"line_number":749,"context_line":"    local ip\u003d$1"},{"line_number":750,"context_line":"    local iface\u003d$(get_if_for_ip \"$ip\")"}],"source_content_type":"application/x-shellscript","patch_set":6,"id":"9a80dd14_c51ec85e","line":747,"in_reply_to":"9a80dd14_32060d1e","updated":"2015-03-07 23:02:45.000000000","message":"Your assumption is wrong. DHCP provides the correct segment MTU to the interface, which wil again reset it.  The path MTU is not guaranteed to be equal to the segment MTU.\n\nNow, granted, in simple cases the path and segment MTU are the same.  But there\u0027s nothing in this patch that acknowledges that, let alone making it known to the devstack user.  Again, you\u0027re solving a subset of the entire problem but you are not making that clear to whoever is using devstack.","commit_id":"d1534734b6e6b9d492efaa1e52a375938ea53184"}],"lib/neutron":[{"author":{"_account_id":3217,"name":"Ian Wells","username":"ijw-ubuntu"},"change_message_id":"957866199b8f8b7e074af23cecb4d9f816d0f104","unresolved":false,"context_lines":[{"line_number":1404,"context_line":"}"},{"line_number":1405,"context_line":""},{"line_number":1406,"context_line":""},{"line_number":1407,"context_line":"function _configure_neutron_dhcp_dnsmask_mtu_options {"},{"line_number":1408,"context_line":"    local mtu\u003d${Q_GUEST_DEVICE_MTU:-$(_neutron_ovs_device_mtu)}"},{"line_number":1409,"context_line":"    if [[ $mtu !\u003d \"nochange\" ]]; then"},{"line_number":1410,"context_line":"        iniset $Q_DHCP_CONF_FILE DEFAULT dnsmasq_config_file \"$NEUTRON_CONF_DIR/dnsmasq.conf\""}],"source_content_type":"application/x-shellscript","patch_set":6,"id":"9a80dd14_d3adc0ea","line":1407,"updated":"2015-03-06 22:16:19.000000000","message":"dnsmasq not dnsmask","commit_id":"d1534734b6e6b9d492efaa1e52a375938ea53184"},{"author":{"_account_id":3217,"name":"Ian Wells","username":"ijw-ubuntu"},"change_message_id":"957866199b8f8b7e074af23cecb4d9f816d0f104","unresolved":false,"context_lines":[{"line_number":1409,"context_line":"    if [[ $mtu !\u003d \"nochange\" ]]; then"},{"line_number":1410,"context_line":"        iniset $Q_DHCP_CONF_FILE DEFAULT dnsmasq_config_file \"$NEUTRON_CONF_DIR/dnsmasq.conf\""},{"line_number":1411,"context_line":"        echo \"dhcp-option\u003d26,$mtu\" \u003e\"$NEUTRON_CONF_DIR/dnsmasq.conf\""},{"line_number":1412,"context_line":"    fi"},{"line_number":1413,"context_line":"}"},{"line_number":1414,"context_line":""},{"line_number":1415,"context_line":"# Restore xtrace"}],"source_content_type":"application/x-shellscript","patch_set":6,"id":"9a80dd14_33eb7c14","line":1412,"updated":"2015-03-06 22:16:19.000000000","message":"Works if and only if all your networks have the same MTU.  If your machine has interfaces with an MTU of 1500, it\u0027s likely your external and provider networks have an MTU of 1500 and any networks created using GRE or VXLAN have an MTU somewhat smaller.  This pretty much guarantees that you\u0027ll get the wrong value.\n\nAlso, this implies that the outward facing interface of routers has a different MTU to the inward facing ones.  I\u0027ve not seen anything in this patch that will make that work correctly - is it addressed?","commit_id":"d1534734b6e6b9d492efaa1e52a375938ea53184"},{"author":{"_account_id":5803,"name":"Attila Fazekas","email":"afazekas@redhat.com","username":"afazekas"},"change_message_id":"c4f42b8ffaa00493bca00af1138e7276b6d85e3e","unresolved":false,"context_lines":[{"line_number":1409,"context_line":"    if [[ $mtu !\u003d \"nochange\" ]]; then"},{"line_number":1410,"context_line":"        iniset $Q_DHCP_CONF_FILE DEFAULT dnsmasq_config_file \"$NEUTRON_CONF_DIR/dnsmasq.conf\""},{"line_number":1411,"context_line":"        echo \"dhcp-option\u003d26,$mtu\" \u003e\"$NEUTRON_CONF_DIR/dnsmasq.conf\""},{"line_number":1412,"context_line":"    fi"},{"line_number":1413,"context_line":"}"},{"line_number":1414,"context_line":""},{"line_number":1415,"context_line":"# Restore xtrace"}],"source_content_type":"application/x-shellscript","patch_set":6,"id":"9a80dd14_df7e9aae","line":1412,"in_reply_to":"9a80dd14_33eb7c14","updated":"2015-03-07 06:38:59.000000000","message":"Your router needs send ask for fragmentation or do the fragmentation, The patch assumes you increased the MTU on your phys dev, because you already did the right thing on the routers/switches.","commit_id":"d1534734b6e6b9d492efaa1e52a375938ea53184"},{"author":{"_account_id":5803,"name":"Attila Fazekas","email":"afazekas@redhat.com","username":"afazekas"},"change_message_id":"950920bdd81fddddfc1fa05be0945ee5b1341ecf","unresolved":false,"context_lines":[{"line_number":1409,"context_line":"    if [[ $mtu !\u003d \"nochange\" ]]; then"},{"line_number":1410,"context_line":"        iniset $Q_DHCP_CONF_FILE DEFAULT dnsmasq_config_file \"$NEUTRON_CONF_DIR/dnsmasq.conf\""},{"line_number":1411,"context_line":"        echo \"dhcp-option\u003d26,$mtu\" \u003e\"$NEUTRON_CONF_DIR/dnsmasq.conf\""},{"line_number":1412,"context_line":"    fi"},{"line_number":1413,"context_line":"}"},{"line_number":1414,"context_line":""},{"line_number":1415,"context_line":"# Restore xtrace"}],"source_content_type":"application/x-shellscript","patch_set":6,"id":"9a80dd14_9d07b817","line":1412,"in_reply_to":"9a80dd14_42ec81f3","updated":"2015-03-09 06:12:22.000000000","message":"s/it can do PMTU discovery/it can assist PMTU discovery/","commit_id":"d1534734b6e6b9d492efaa1e52a375938ea53184"},{"author":{"_account_id":3217,"name":"Ian Wells","username":"ijw-ubuntu"},"change_message_id":"f0aec1c81fe036b741e7f8e6a72c837b659419c3","unresolved":false,"context_lines":[{"line_number":1409,"context_line":"    if [[ $mtu !\u003d \"nochange\" ]]; then"},{"line_number":1410,"context_line":"        iniset $Q_DHCP_CONF_FILE DEFAULT dnsmasq_config_file \"$NEUTRON_CONF_DIR/dnsmasq.conf\""},{"line_number":1411,"context_line":"        echo \"dhcp-option\u003d26,$mtu\" \u003e\"$NEUTRON_CONF_DIR/dnsmasq.conf\""},{"line_number":1412,"context_line":"    fi"},{"line_number":1413,"context_line":"}"},{"line_number":1414,"context_line":""},{"line_number":1415,"context_line":"# Restore xtrace"}],"source_content_type":"application/x-shellscript","patch_set":6,"id":"9a80dd14_e57f4c64","line":1412,"in_reply_to":"9a80dd14_df7e9aae","updated":"2015-03-07 23:02:45.000000000","message":"To give an example: let\u0027s say I have one interface which serves at the interface over which tunnels are carried and also, independently, the router external network, and it has an MTU of 2000 set on the interface.  (If this doesn\u0027t make you happy you can also consider a 2-interface case with the same MTU on both interfaces; it works the same.)\n\nInterfaces on tenant virtual networks will have an MTU of (2000 - the encap size).\nInterfaces on the external network should have an MTU of 2000 (which is also what non-Openstack devices on that network will have configured as their MTU).\n\nIf a VM or router is connected to the external network and its MTU is *not* 2000, it will fail to communicate with its upstream router - some packets will be sent from one side to the other that are too large to be received, and will be dropped.  For instance, your patch will calculate an MTU of 1950 (or whatever) based on the encap and set it for all networks.  The upstream interface of neutron routers, connected to the external network, will have an MTU of 1950. The next machine upstream of Openstack has an MTU set of 2000, and will occasionally transmit packets of more than 1950 bytes to the Neutron routers on the Openstack external network based on the MTU set on its interface.  Openstack will drop them because they are bigger than the 1950 MTU of its routers.\n\nThe upstream router *will* fragment packets, absolutely.  However, it will only do it if its MTU indicates that it has to do it, and in this case it doesn\u0027t.","commit_id":"d1534734b6e6b9d492efaa1e52a375938ea53184"},{"author":{"_account_id":5803,"name":"Attila Fazekas","email":"afazekas@redhat.com","username":"afazekas"},"change_message_id":"887aa89fb74c8339adf385429a6a3c1f5db15e53","unresolved":false,"context_lines":[{"line_number":1409,"context_line":"    if [[ $mtu !\u003d \"nochange\" ]]; then"},{"line_number":1410,"context_line":"        iniset $Q_DHCP_CONF_FILE DEFAULT dnsmasq_config_file \"$NEUTRON_CONF_DIR/dnsmasq.conf\""},{"line_number":1411,"context_line":"        echo \"dhcp-option\u003d26,$mtu\" \u003e\"$NEUTRON_CONF_DIR/dnsmasq.conf\""},{"line_number":1412,"context_line":"    fi"},{"line_number":1413,"context_line":"}"},{"line_number":1414,"context_line":""},{"line_number":1415,"context_line":"# Restore xtrace"}],"source_content_type":"application/x-shellscript","patch_set":6,"id":"9a80dd14_42ec81f3","line":1412,"in_reply_to":"9a80dd14_e57f4c64","updated":"2015-03-09 05:34:40.000000000","message":"The physical router knows all interfaces segment MTU, and it can do PMTU discovery.\nThe worst thing what a physical router can do is silently drop the frame, in this case the router is miss configured.\n\nUnfortunately OVS does silent drops, that\u0027s why we need to configure mtus.\n\nI did not added br-ex mtu yet :(\nhttps://github.com/openstack-infra/devstack-gate/blob/master/devstack-vm-gate.sh#L532\n\nThis is what you actually miss ?","commit_id":"d1534734b6e6b9d492efaa1e52a375938ea53184"}],"lib/neutron_plugins/openvswitch_agent":[{"author":{"_account_id":970,"name":"Dean Troyer","email":"dtroyer@gmail.com","username":"dtroyer"},"change_message_id":"1e071ba4faa592b242b41934a799ed1fd55cb377","unresolved":false,"context_lines":[{"line_number":28,"context_line":"function _configure_dhcp_options {"},{"line_number":29,"context_line":"    local mtu\u003d${GUEST_DEVICE_MTU:-$(_neutron_ovs_device_mtu)}"},{"line_number":30,"context_line":"    if [[ $mtu !\u003d \"nochange\" ]]; then"},{"line_number":31,"context_line":"        iniset $Q_DHCP_CONF_FILE DEFAULT dnsmasq_config_file \"/etc/neutron/dnsmasq.conf\""},{"line_number":32,"context_line":"        echo \"dhcp-option\u003d26,$mtu\" \u003e/etc/neutron/dnsmasq.conf"},{"line_number":33,"context_line":"    fi"},{"line_number":34,"context_line":"}"}],"source_content_type":"application/octet-stream","patch_set":1,"id":"baada198_55f9cd4d","line":31,"updated":"2014-08-08 04:19:52.000000000","message":"These paths should use NEUTRON_CONF_DIR","commit_id":"3340fd4af2bc4b325eaa8c40a4d203e83325bc0e"},{"author":{"_account_id":970,"name":"Dean Troyer","email":"dtroyer@gmail.com","username":"dtroyer"},"change_message_id":"00165609e1decae4dd1d82ec21050c323d21d0e1","unresolved":false,"context_lines":[{"line_number":29,"context_line":"    local mtu\u003d${GUEST_DEVICE_MTU:-$(_neutron_ovs_device_mtu)}"},{"line_number":30,"context_line":"    if [[ $mtu !\u003d \"nochange\" ]]; then"},{"line_number":31,"context_line":"        iniset $Q_DHCP_CONF_FILE DEFAULT dnsmasq_config_file \"$NEUTRON_CONF_DIR/dnsmasq.conf\""},{"line_number":32,"context_line":"        echo \"dhcp-option\u003d26,$mtu\" \u003e/etc/neutron/dnsmasq.conf"},{"line_number":33,"context_line":"    fi"},{"line_number":34,"context_line":"}"},{"line_number":35,"context_line":""}],"source_content_type":"application/octet-stream","patch_set":2,"id":"baada198_6ffdb907","line":32,"updated":"2014-08-12 03:35:20.000000000","message":"NEUTRON_CONF_DIR here too?","commit_id":"618c4314f6da9fa72b7dd9153263f4cbf85293a6"},{"author":{"_account_id":7118,"name":"Ian Wienand","email":"iwienand@redhat.com","username":"iwienand"},"change_message_id":"91918228f5420179659657ac3702678709d83253","unresolved":false,"context_lines":[{"line_number":26,"context_line":"}"},{"line_number":27,"context_line":""},{"line_number":28,"context_line":"function _configure_dhcp_options {"},{"line_number":29,"context_line":"    local mtu\u003d${GUEST_DEVICE_MTU:-$(_neutron_ovs_device_mtu)}"},{"line_number":30,"context_line":"    if [[ $mtu !\u003d \"nochange\" ]]; then"},{"line_number":31,"context_line":"        iniset $Q_DHCP_CONF_FILE DEFAULT dnsmasq_config_file \"$NEUTRON_CONF_DIR/dnsmasq.conf\""},{"line_number":32,"context_line":"        echo \"dhcp-option\u003d26,$mtu\" \u003e\"$NEUTRON_CONF_DIR/dnsmasq.conf\""}],"source_content_type":"application/octet-stream","patch_set":3,"id":"baada198_c71cc8c6","line":29,"updated":"2014-08-12 20:14:39.000000000","message":"why not have this in lib/neutron as a global so it is discoverble\n\nit also seems like it should be namespaced to be more specific","commit_id":"ed5810d011e7d1ee98653032c46687acf73fe9cd"},{"author":{"_account_id":3217,"name":"Ian Wells","username":"ijw-ubuntu"},"change_message_id":"957866199b8f8b7e074af23cecb4d9f816d0f104","unresolved":false,"context_lines":[{"line_number":29,"context_line":""},{"line_number":30,"context_line":"function neutron_plugin_configure_dhcp_agent {"},{"line_number":31,"context_line":"    iniset $Q_DHCP_CONF_FILE DEFAULT dhcp_agent_manager neutron.agent.dhcp_agent.DhcpAgentWithStateReport"},{"line_number":32,"context_line":"    _configure_neutron_dhcp_dnsmask_mtu_options"},{"line_number":33,"context_line":"}"},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"function neutron_plugin_configure_l3_agent {"}],"source_content_type":"application/x-shellscript","patch_set":6,"id":"9a80dd14_d3db007a","line":32,"updated":"2015-03-06 22:16:19.000000000","message":"dnsmasq not dnsmask","commit_id":"d1534734b6e6b9d492efaa1e52a375938ea53184"}],"lib/neutron_plugins/ovs_base":[{"author":{"_account_id":970,"name":"Dean Troyer","email":"dtroyer@gmail.com","username":"dtroyer"},"change_message_id":"1e071ba4faa592b242b41934a799ed1fd55cb377","unresolved":false,"context_lines":[{"line_number":103,"context_line":"                (( mtu-\u003d 28 ))"},{"line_number":104,"context_line":"                ;;"},{"line_number":105,"context_line":"            *)"},{"line_number":106,"context_line":"                echo \"Warning: the optimal device MTU is unknown for your tunneling options!\"\u003e2"},{"line_number":107,"context_line":"                mtu\u003d\"nochange\""},{"line_number":108,"context_line":"                ;;"},{"line_number":109,"context_line":"        esac"}],"source_content_type":"application/octet-stream","patch_set":1,"id":"baada198_d5623d14","line":106,"updated":"2014-08-08 04:19:52.000000000","message":"We\u0027ve done unspeakable things to stdout and stderr for logging purposes.  Use the warn() function for things like this:\n\n  warn $LINENO \"The optimal device...\"","commit_id":"3340fd4af2bc4b325eaa8c40a4d203e83325bc0e"},{"author":{"_account_id":5803,"name":"Attila Fazekas","email":"afazekas@redhat.com","username":"afazekas"},"change_message_id":"950920bdd81fddddfc1fa05be0945ee5b1341ecf","unresolved":false,"context_lines":[{"line_number":101,"context_line":"    fi"},{"line_number":102,"context_line":"    local mtu\u003d${Q_GUEST_DEVICE_MTU:-$(_neutron_ovs_device_mtu)}"},{"line_number":103,"context_line":"    if [[ $mtu !\u003d \"nochange\" ]]; then"},{"line_number":104,"context_line":"        sudo ip link set mtu $mtu dev $PUBLIC_BRIDGE"},{"line_number":105,"context_line":"    fi"},{"line_number":106,"context_line":"}"},{"line_number":107,"context_line":""}],"source_content_type":"application/x-shellscript","patch_set":6,"id":"9a80dd14_020b692a","line":104,"updated":"2015-03-09 06:12:22.000000000","message":"Actually I did it.","commit_id":"d1534734b6e6b9d492efaa1e52a375938ea53184"},{"author":{"_account_id":3217,"name":"Ian Wells","username":"ijw-ubuntu"},"change_message_id":"957866199b8f8b7e074af23cecb4d9f816d0f104","unresolved":false,"context_lines":[{"line_number":110,"context_line":"# because many of networking operation is done per frame basis."},{"line_number":111,"context_line":"# If you can send the same amount of data in less frames you need to handle less IRQ,"},{"line_number":112,"context_line":"# process less packet header and you are using less bytes for protocol headers in total."},{"line_number":113,"context_line":"# This function assumes you already configured your physical interface with the maximum MTU,"},{"line_number":114,"context_line":"# which can be handled by your NIC and physical switches. If you would like to use different MTU"},{"line_number":115,"context_line":"# then the detected one use ``Q_GUEST_DEVICE_MTU`` variable."},{"line_number":116,"context_line":"# By default the guests assumes 1500 byte MTU, which can be problematic when you are using"}],"source_content_type":"application/x-shellscript","patch_set":6,"id":"9a80dd14_93cd08ba","line":113,"updated":"2015-03-06 22:16:19.000000000","message":"Segment MTU is not the same as path MTU.  The interface\u0027s confgured MTU is for the L2 segment it\u0027s on.  It\u0027s possible that the maximum size of unfragmented packet it can get to any specific IP (for instance, another compute host) is smaller than this size.  \n\nThus this approach will work for people except when it doesn\u0027t - it\u0027s far from a comprehensive fix.  I\u0027ve not seen anything in the comments that makes this clear, and I wouldn\u0027t necessarily expect your average devstack user to be reading the comments anyway.","commit_id":"d1534734b6e6b9d492efaa1e52a375938ea53184"},{"author":{"_account_id":5803,"name":"Attila Fazekas","email":"afazekas@redhat.com","username":"afazekas"},"change_message_id":"c4f42b8ffaa00493bca00af1138e7276b6d85e3e","unresolved":false,"context_lines":[{"line_number":110,"context_line":"# because many of networking operation is done per frame basis."},{"line_number":111,"context_line":"# If you can send the same amount of data in less frames you need to handle less IRQ,"},{"line_number":112,"context_line":"# process less packet header and you are using less bytes for protocol headers in total."},{"line_number":113,"context_line":"# This function assumes you already configured your physical interface with the maximum MTU,"},{"line_number":114,"context_line":"# which can be handled by your NIC and physical switches. If you would like to use different MTU"},{"line_number":115,"context_line":"# then the detected one use ``Q_GUEST_DEVICE_MTU`` variable."},{"line_number":116,"context_line":"# By default the guests assumes 1500 byte MTU, which can be problematic when you are using"}],"source_content_type":"application/x-shellscript","patch_set":6,"id":"9a80dd14_ff527e33","line":113,"in_reply_to":"9a80dd14_93cd08ba","updated":"2015-03-07 06:38:59.000000000","message":"It is possible go wrong in a very heterogeneous network, but in this case most likely you wouldn\u0027t changed the MTU on your phys dev to an incompatible value.","commit_id":"d1534734b6e6b9d492efaa1e52a375938ea53184"},{"author":{"_account_id":3217,"name":"Ian Wells","username":"ijw-ubuntu"},"change_message_id":"f0aec1c81fe036b741e7f8e6a72c837b659419c3","unresolved":false,"context_lines":[{"line_number":110,"context_line":"# because many of networking operation is done per frame basis."},{"line_number":111,"context_line":"# If you can send the same amount of data in less frames you need to handle less IRQ,"},{"line_number":112,"context_line":"# process less packet header and you are using less bytes for protocol headers in total."},{"line_number":113,"context_line":"# This function assumes you already configured your physical interface with the maximum MTU,"},{"line_number":114,"context_line":"# which can be handled by your NIC and physical switches. If you would like to use different MTU"},{"line_number":115,"context_line":"# then the detected one use ``Q_GUEST_DEVICE_MTU`` variable."},{"line_number":116,"context_line":"# By default the guests assumes 1500 byte MTU, which can be problematic when you are using"}],"source_content_type":"application/x-shellscript","patch_set":6,"id":"9a80dd14_2595e486","line":113,"in_reply_to":"9a80dd14_ff527e33","updated":"2015-03-07 23:02:45.000000000","message":"This has nothing to do with the value on the physical device (which is correct so long as it agrees with the network segment it\u0027s attached to) but I agree it\u0027s a rare case.  My issue is that there\u0027s no indication to the devstack user that it won\u0027t work or what to do about it.","commit_id":"d1534734b6e6b9d492efaa1e52a375938ea53184"},{"author":{"_account_id":6698,"name":"Robert Pothier","email":"robpothier35@gmail.com","username":"rpothier"},"change_message_id":"01a0c87536e65dab8358533cf7da47592672684e","unresolved":false,"context_lines":[{"line_number":161,"context_line":"    if [[ $mtu !\u003d \"nochange\" ]]; then"},{"line_number":162,"context_line":"        # use the same value in both neutron.conf and nova.conf"},{"line_number":163,"context_line":"        iniset $NOVA_CONF DEFAULT network_device_mtu \"$mtu\""},{"line_number":164,"context_line":"        iniset $NEUTRON_CONF DEFAULT network_device_mtu \"$mtu\""},{"line_number":165,"context_line":"        iniset /$Q_PLUGIN_CONF_FILE agent veth_mtu $((mtu + 4))"},{"line_number":166,"context_line":"    fi"},{"line_number":167,"context_line":"}"}],"source_content_type":"application/x-shellscript","patch_set":6,"id":"9a80dd14_523f18d6","line":164,"updated":"2015-03-06 16:43:49.000000000","message":"for neutron, can it just use Q_GUEST_DEVICE_MTU to set the the segment mtu, without subtracting the protocol.","commit_id":"d1534734b6e6b9d492efaa1e52a375938ea53184"}]}