Network Working Group Danny McPherson INTERNET DRAFT Amber Networks, Inc. February 2001 IS-IS Transient Blackhole Avoidance 1. Status of this Memo This document is an Internet-Draft and is in full conformance with all provisions of Section 10 of RFC 2026. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet- Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet- Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html. 2. Abstract This document describes a simple, interoperable mechanism that can be employed in IS-IS networks in order to decrease data loss associated with deterministic blackholing of packets during transient network conditions. The mechanism proposed here requires no IS-IS protocol changes and is completely interoperable with the existing IS-IS specification. McPherson, D. [Page 1] INTERNET DRAFT February 2001 3. Introduction When an IS-IS router that was previously a transit router becomes unavailable as a result of some transient condition such as a reboot, other routers within the routing domain must select an alternative path to reach destinations which had previously transited the failed router. Presumably, the newly selected router(s) comprising the path have been available for some time and, as a result, have complete forwarding information bases (FIBs) which contain a full set of reachability information for both internal and external (e.g., BGP) destination networks. When the previously failed router becomes available again, in only a few seconds paths that had previously transited the router are again selected as the optimal path by the IGP. As a result, forwarding tables are updated and packets are once again forwarded along the path. Unfortunately, external destination reachability information (e.g., learned via BGP) is not yet available to the router, and as a result, packets bound for destinations not learned via the IGP are unnecessarily discarded. A simple interoperable mechanism to alleviate the offshoot associated with this deterministic behavior is discussed below. 4. Discussion This document describes a simple, interoperable mechanism that can be employed in IS-IS [1, 2] networks in order to avoid transition to a newly available path until other associated routing protocols such as BGP have had sufficient time to converge. The benefits of such a mechanism can realized when considering the following scenario. McPherson, D. [Page 2] INTERNET DRAFT February 2001 D.1 | +-------+ | RtrD | +-------+ / \ / \ +-------+ +-------+ | RtrB | | RtrC | +-------+ +-------+ \ / \ / +-------+ | RtrA | +-------+ | S.1 Host S.1 is transmitting data to destination D.1 via a primary path of RtrA->RtrB->RtrD. Routers A, B and C learn of reachability to destination D.1 via BGP from RtrD. RtrA's primary path to D.1 is selected because when calculating the path to BGP NEXT_HOP of RtrD the sum of the IS-IS link metrics on the RtrA-RtrB-RtrD path is less than the sum of the metrics of the RtrA-RtrC-RtrD path. Assume RtrB becomes unavailable and as a result the RtrC path to RtrD is used. Once RtrA's FIB is updated and it begins forwarding packets to RtrC everything should behave properly as RtrC has existing for=AD warding information regarding destination D.1's availability via BGP NEXT_HOP RtrD. Assume now that RtrB comes back online. In only a few seconds IS-IS neighbor state has been established with RtrA and RtrD and database synchronization has occurred. RtrA now realizes that the best path to destination D.1 is via RtrB, and therefore updates it FIB appro=AD priately. RtrA begins to forward packets destined for D.1 to RtrB. Though, because RtrB has yet to establish and synchronize it's BGP neighbor relationship and routing information with RtrD, RtrB has no knowledge regarding reachability of destination D.1, and therefore discards the packets received from RtrA destined to D.1. If RtrB were to temporarily set it's LSP Overload bit while synchro=AD= nizing BGP tables with it's neighbors, RtrA would continue to use the working RtrA->RtrC->RtrD path, and the LSP should only be used to obtain reachability to locally connected networks (rather than for calculating transit paths through the router, as defined in [1]). McPherson, D. [Page 3] INTERNET DRAFT February 2001 After initial synchronization of BGP tables with neighboring routers, RtrB would generate a new LSP, clearing the Overload bit, and RtrA could again begin using the optimal path via RtrB. Typically, in service provider networks IBGP connections are done via peerings with 'loopback' addresses. As such, the newly available router must advertise it's own loopback (or similar) IP address, as well as associated adjacencies, in order to make the loopbacks acces=AD= sible to other routers within the routing domain. It's because of this that simply flooding an empty LSP is not sufficient. 5. Deployment Considerations Such a mechanism increases overall network availability and allows network operators to alleviate the deterministic blackholing behavior introduced in this scenario. Similar mechanisms [3] have been defined for OSPF, though only after realizing usefulness obtained from that of the IS-IS Overload bit. This mechanism has been deployed in several large IS-IS networks for several of years. Triggers for setting the Overload bit as described are left to the implementer. Some potential triggers could perhaps include "N sec=AD onds after booting", or "N number of BGP prefixes in the BGP Loc- RIB". Unlike similar mechanisms employed in [3], if the Overload bit is set in a router's LSP, NO transit paths are calculated through the router. As such, if no alternative paths are available to the desti=AD= nation network, employing such a mechanism may actually have a nega=AD= tive impact on convergence (i.e., the router maintains the only available path to reach downstream routers, but the OL bit disallows other nodes in the network from calculating paths via the router, and as such, no feasible path exists to the routers). Finally, if all systems within an IS-IS routing domain have NOT implemented the Overload bit correctly, forwarding loops may occur. McPherson, D. [Page 4] INTERNET DRAFT February 2001 6. Potential Alternatives Alternatively, it may be considered more appealing to employ some=AD thing more akin to [3] for this purpose. With this model, during transient conditions a node advertises excessively high link metrics to serve as an indication to other nodes in the network that paths transiting the router are less desirable than existing paths. The advantage of a metric-based mechanism over the OL bit mechanism proposed here model is that transit paths may still be calculated through the router. Another advantage is that a metric-based mecha=AD= nism does not require that all nodes in the IS-IS domain correctly implement the OL bit. However, as currently deployed, IS-IS provides for only 6 bits of space for link metric allocation, and 10 bits aggregate path metric. Though extensions proposed in [4] remove this limitation, they've not yet been widely deployed. As such, there's currently little flexi=AD bility when using link metrics for this purpose. Of course, both methods proposed in this document are backwards-com=AD= patible and implementation-specific. 7. Security Considerations The mechanisms specified in this memo introduces no new security issues to IS-IS. 8. Acknowledgments The author of this document makes no claim to the originality of the idea. Others To be supplied... 9. References [1] ISO, "Intermediate system to Intermediate system routeing information exchange protocol for use in conjunction with the Protocol for providing the Connectionless-mode Network Service (ISO 8473)," ISO/IEC 10589:1992. [2] Callon, R., "OSI IS-IS for IP and Dual Environment," RFC 1195, December 1990. [3] Retana et al., "OSPF Stub Router Advertisement", "Work in Progress", November 2000. McPherson, D. [Page 5] INTERNET DRAFT February 2001 [4] Li, T., Smit, H., "IS-IS extensions for Traffic Engineering", Work in Progress. 10. Authors' Address Danny McPherson Amber Networks, Inc. 48664 Milmont Drive Fremont, CA 94538 Phone: 510.687.5200 Email: danny@ambernetworks.com McPherson, D. [Page 6]