From 2aa200bdb994a72e8159f78c4439b81232436942 Mon Sep 17 00:00:00 2001 From: Slawek Kaplonski Date: Wed, 10 Apr 2019 12:49:49 +0200 Subject: [PATCH] Choose random value for HA routes' vr_id HA routers are using keepalived and needs to have virtual_router_id configured. As routers which belongs to same tenant are using same ha network, those values have to be different for each router. Before this patch this value was always taken as first available value from available_vr_ids range. In some (rare) cases, when more than one router is created in parallel for same tenant it may happen that those routers would have same vr_id choosen so keepalived would treat them as single application and only one router would be ACTIVE on one of L3 agents. This patch changes this behaviour that now random value from available vr_ids will be chosen instead of taking first value always. That should mittigate this rare race condition that it will be (almost) not noticable for users. However, proper fix should be probably done as some additional constraint in database layer. But such solution wouldn't be possible to backport to stable branches so I decided to propose this easy patch first. Change-Id: Idb0ed744e54976dca23593fb2d7317bf77442e65 Related-Bug: #1823314 (cherry picked from commit a8d0f557d504957aeb91f451105cca9eee2d6adb) --- neutron/db/l3_hamode_db.py | 21 ++++++++++++++------- 1 file changed, 14 insertions(+), 7 deletions(-) diff --git a/neutron/db/l3_hamode_db.py b/neutron/db/l3_hamode_db.py index 09af7d5c41d..01a5f71871c 100644 --- a/neutron/db/l3_hamode_db.py +++ b/neutron/db/l3_hamode_db.py @@ -14,6 +14,7 @@ # import functools +import random import netaddr from neutron_lib.api.definitions import l3 as l3_apidef @@ -105,6 +106,16 @@ class L3_HA_NAT_db_mixin(l3_dvr_db.L3_NAT_with_dvr_db_mixin, allocated_vr_ids = set(a.vr_id for a in vr_id_objs) - set([0]) return allocated_vr_ids + def _get_vr_id(self, context, network_id): + allocated_vr_ids = self._get_allocated_vr_id(context, + network_id) + available_vr_ids = VR_ID_RANGE - allocated_vr_ids + + if not available_vr_ids: + return None + + return random.choice(list(available_vr_ids)) + @db_api.retry_if_session_inactive() def _ensure_vr_id(self, context, router_db, ha_network): router_id = router_db.id @@ -126,16 +137,12 @@ class L3_HA_NAT_db_mixin(l3_dvr_db.L3_NAT_with_dvr_db_mixin, return old_router = self._make_router_dict(router_db) - allocated_vr_ids = self._get_allocated_vr_id(context, - network_id) - available_vr_ids = VR_ID_RANGE - allocated_vr_ids - - if not available_vr_ids: + vr_id = self._get_vr_id(context, network_id) + if vr_id is None: raise l3ha_exc.NoVRIDAvailable(router_id=router_id) allocation = l3_hamode.L3HARouterVRIdAllocation( - context, network_id=network_id, - vr_id=available_vr_ids.pop()) + context, network_id=network_id, vr_id=vr_id) allocation.create() router_db.extra_attributes.ha_vr_id = allocation.vr_id