This document is intended to present a structured troubleshooting methodology and useful tools to help identify and isolate Group Encrypted Transport VPN (GETVPN) problems and to provide possible solutions.
Cisco recommends that you have knowledge of these topics:
This document is not restricted to specific software and hardware versions.
The information in this document was created from the devices in a specific lab environment. All of the devices used in this document started with a cleared (default) configuration. If your network is live, make sure that you understand the potential impact of any command.
GETVPN Troubleshooting Methodology
As with most troubleshooting of complex technology problems, the key is to be able to isolate the problem to a specific feature, subsystem, or component. The GETVPN solution is comprised of a number of feature components, specifically:
Internet Key Exchange (IKE) - Used between Group Member (GM) and Key Server (KS), and amongst Cooperative Protocol (COOP) KSs in order to authenticate and protect the Control Plane.
Group Domain of Interpretation (GDOI) - Protocol used for the KS in order to distribute group keys and provide key service such as rekey to all the GMs.
COOP - Protocol used for the KSs in order to communicate with each other and provide redundancy.
Header Preservation - IPsec in Tunnel mode that preserves the original data packet header for end-to-end traffic delivery.
Time Based Anti-Replay (TBAR) - Replay detection mechanism used in a group key environment.
It also provides an extensive set of troubleshooting tools in order to ease the troubleshoot process. It is important to understand which of these tools are available, and when they are appropriate for each troubleshooting task. When troubleshooting, it is always a good idea to start with the least intrusive methods so that the production environment is not negatively impacted. The key to this structured troubleshooting is to be able to break the problem down to either a control or data plane issue. You can do this if you follow the protocol or data flow and use the various tools presented here in order to checkpoint them.
This GETVPN topology and addressing scheme is used throughout the rest of this troubleshooting document.
crypto gdoi group G1 identity number 3333 server local rekey authenmypubkeyrsa get rekey transport unicast sa ipsec 1 profile gdoi-p match address ipv4ENCPOL address ipv4 10.1.11.2 redundancy local priority 10 peer address ipv4 10.1.12.2
crypto gdoi group G1 identity number 3333 server address ipv4 10.1.11.2 server address ipv4 10.1.12.2 ! crypto map gm_map 10 gdoi set group G1 ! interface Serial1/0 crypto map gm_map
Note: The KS2 and GM2 configurations are not included here for brevity.
KS - Key Server
GM - Group Member
COOP - Cooperative Protocol
TBAR - Time Based Anti-Replay
KEK - Key Encryption Key
TEK - Traffic Encryption Key
Logging Facility Preparation and Other Best Practices
Before you begin to troubleshoot, ensure that you have prepared the logging facility as described here. Some best practices are also listed here:
Check the router amount of free memory, and configure logging buffered debugging to a large value (10 MB or more if possible).
Disable logging to the console, monitor, and syslog servers.
Retrieve the logging buffer content with the show log command at regular intervals, every 20 mins to an hour, in order to prevent log loss due to buffer reuse.
Whatever happens, enter the show tech command from affected GMs and KSs, and examine the output of the show ip route command in global and each Virtual Routing and Forwarding (VRF) involved, if any are required.
Use Network Time Protocol (NTP) in order to sync the clock between all devices that are debugged. Enable millisecond (msec) timestamps for both debug and log messages:
service timestamps debug datetime msec service timestamps log datetime msec
Make sure the show command outputs are timestamped.
Router#terminal exec prompt timestamp
When you collect show command outputs for control plane events or data plane counters, always collect multiple iterations of the same output.
Troubleshoot GETVPN Control Plane Issues
Control plane means all the protocol events that led up to the policy and Security Association (SA) creation on the GM so that they are ready to encrypt and decrypt data plane traffic. Some of the key checkpoints in the GETVPN control plane are:
Control Plane Debugging Best Practices
These troubleshooting best practices are not GETVPN specific; they apply to almost any control plane debugging. It is critical to follow these best practices in order to ensure the most effective troubleshooting:
Turn off console logging and use the logging buffer or syslog in order to collect the debugs.
Use NTP in order to sync router clocks on all the devices that are debugged.
Enable msec timestamping for debug and log messages:
service timestamp debug datetime msec service timestamp log datetime msec
Make sure the show command outputs are timestamped so that they can be correlated with the debug output:
terminal exec prompt timestamp
Use conditional debugging in a scale environment if possible.
GETVPN Control Plane Troubleshooting Tools
GETVPN Show Commands
As a general rule, these are the command outputs you should collect for almost all GETVPN problems.
show crypto gdoi show crypto gdoi ks coop show crypto gdoi ks members show crypto gdoi ks rekey show crypto gdoi ks policy
show crypto eli show crypto gdoi rekey sa show crypto gdoi show crypto gdoi gm show crypto gdoi gm rekey
GETVPN Syslog Messages
GETVPN provides an extensive set of syslog messages for significant protocol events and error conditions. The syslog should always be the first place to look when you perform GETVPN troubleshooting.
Common KS Syslog Messages
The configuration between the primary key server and secondary key server is mismatched.
The local key server has entered the election process in a group.
The reachability between the configured cooperative key servers is restored.
The local key server transitioned to a primary role from being a secondary server in a group.
An authorized remote server tried to contact the local key server in a group, which could be considered a hostile event.
The reachability between the configured cooperative key servers is lost, which could be considered a hostile event.
During rekey protocol, an unauthorized member tried to join a group, which could be considered a hostile event.
Sending multicast rekey.
Sending unicast rekey.
During GDOI registration protocol, an unauthorized member tried to join a group, which could be considered a hostile event.
The registration request was dropped because the requesting device was not authorized to join the group.
Common GM Syslog Messages
The clear crypto gdoi command has been executed by the local group member.
A crypto map has been attached for the local group member.
A crypto map has been detached for the local group member.&
IPsec SA created for one group might have been expired or cleared. Need to reregister to the key server.
Group member has transitioned from using a unicast rekey mechanism to using a multicast mechanism.
Group member has transitioned from using a multicast rekey mechanism to using a unicast mechanism.
A group member has received a pseudotime with a value that is largely different from its own pseudotime.
A group member or key server has failed an anti-replay check.
Note: The messages highlighted in red are the most common or significant messages seen in a GETVPN environment.
Global Crypto and GDOI Debugs
GETVPN debugs are divided:
First by the device on which you are troubleshooting.
F340.06.15-2900-18#debug cry gdoi ? all-features All features in GDOI condition GDOI Conditional Debugging gm Group Member ks Key Server
Second by the type of problem you are troubleshooting.
GM1#debug cry gdoi gm ? all-features All Group Member features infrastructure GM Infrastructure registration GM messages related to registration rekey GM messagese related to Re-Key replay Anti Replay
Third by the level of debugging that needs to be enabled. In Version 15.1(3)T and later, all GDOI feature debugs were standardized to have these debug levels. This was designed in order to help troubleshoot large-scale GETVPN environments with enough debugging granularity. When you debug GETVPN problems, it is important to use the appropriate debug level. As a general rule, start with the lowest debug level, that is the error level, and increase the debugging granularity when needed.
Important messages to the user and protocol issues
State transitions and events such as send and receive rekeys
Most detailed debug message information
Includes dump of detailed packet information
All of the above
GDOI Conditional Debugging
In Cisco IOS® Version 15.1(3)T and later, GDOI conditional debugging was added in order to help troubleshoot GETVPN in a large-scale environment. So all Internet Security Association and Key Management Protocol (ISAKMP) and GDOI debugs can now be triggered with a conditional filter based on the group or peer IP address. For most GETVPN problems, it is good to enable both ISAKMP and GDOI debugs with the appropriate conditional filter, since GDOI debugs only show GDOI-specific operations. In order to use ISAKMP and GDOI conditional debugs, complete these two simple steps:
Note: With both ISAKMP and GDOI conditional debugs, in order to catch debug messages that might not have the conditional filter information, for example the IP address in the debug path, the unmatched flag can be enabled. However, this must be used with caution because it can produce a large amount of debug information.
GDOI Event Traces
This was added in Version 15.1(3)T. Event tracing offers light-weight, always-on tracing for significant GDOI events and errors. There is also exit-path tracing with traceback enabled for exception conditions. Event traces can provide more GETVPN event history information than traditional syslogs.
GDOI event traces are enabled by default and can be retrieved from the trace buffer with the show monitor even-trace command.
GM1#show monitor event-trace gdoi ? all Show all the traces in current buffer back Show trace from this far back in the past clock Show trace from a specific clock time/date coop GDOI COOP Event Traces exit GDOI Exit Traces from-boot Show trace from this many seconds after booting infra GDOI INFRA Event Traces latest Show latest trace events since last display merged Show entries in all event traces sorted by time registration GDOI Registration event Traces rekey GDOI Rekey event Traces
GM1#show monitor event-trace gdoi rekey all *Nov 6 15:55:16.117: GDOI_REKEY_EVENT: ACK_SENT: From 10.1.12.2 to 10.1.13.2 with seq no 1 for the group G1 *Nov 6 15:55:16.117: GDOI_REKEY_EVENT: REKEY_RCVD: From 10.1.12.2 to 10.1.13.2 with seq no 1 for the group G1 *Nov 6 16:11:01.125: GDOI_REKEY_EVENT: ACK_SENT: From 10.1.12.2 to 10.1.13.2 with seq no 1 for the group G1 *Nov 6 16:11:01.125: GDOI_REKEY_EVENT: REKEY_RCVD: From 10.1.12.2 to 10.1.13.2 with seq no 1 for the group G1
The exit path trace provides detailed information about exit path, that is exception and error conditions, with the traceback option enabled by default. The tracebacks can then be used in order to decode the exact code sequence that has led to the exit path condition. Use the detail option in order to retrieve the tracebacks from the trace buffer:
The default trace buffer size is 512 entries, and this might not be enough if the problem is intermittent. In order to increase this default trace entry size, the event trace configuration parameters can be changed like shown here:
GM1#show monitor event-trace gdoi rekey parameters Trace has 512 entries Stacktrace is disabled by default
GM1# GM1#config t Enter configuration commands, one per line. End with CNTL/Z. GM1(config)#monitor event-trace gdoi rekey size ? <1-1000000> Number of entries in trace
GETVPN Control Plane Checkpoints and Common Issues
Here are some of the common control plane issues for GETVPN. To re-iterate, the Control Plane is defined as all of the GETVPN feature components required in order to enable dataplane encryption and decryption on the GMs. At a high level, this requires successful GM registration, security policy and SA download/install, and subsequent KEK/TEK rekey.
COOP Setup and Policy Creation
In order to check and verify that the KS has successfully created the security policy and the associated KEK/TEK, enter:
KS1#show crypto gdoi ks policy Key Server Policy: For group G1 (handle: 2147483650) server 10.1.11.2 (handle: 2147483650):
For group G1 (handle: 2147483650) server 10.1.12.2 (handle: 2147483651):
# of teks : 1 Seq num : 10 KEK POLICY (transport type : Unicast) spi : 0x18864836BA888BCD1126671EEAFEB4C7 management alg : disabled encrypt alg : 3DES crypto iv length : 8 key size : 24 orig life(sec): 1200 remaining life(sec): 528 sig hash algorithm : enabled sig key length : 162 sig size : 128 sig key name : key1
One common problem with the KS policy setup is when there are different policies configured between the primary and secondary KSs. This can result in unpredictable KS behavior and this error will be reported:
%GDOI-3-COOP_CONFIG_MISMATCH: WARNNING: replay method configuration between Primary KS and Secondary KS are mismatched
Currently there is no automatic configuration sync between primary and secondary KSs, so these must be manually rectified.
Because COOP is a critical (and almost always mandatory) configuration for GETVPN, it is key to make sure COOP works correctly and the COOP KS roles are correct:
KS1#show crypto gdoi ks coop Crypto Gdoi Group Name :G1 Group handle: 2147483650, Local Key Server handle: 2147483650
Local Address: 10.1.11.2 Local Priority: 200 Local KS Role: Primary , Local KS Status: Alive Local KS version: 1.0.4 Primary Timers: Primary Refresh Policy Time: 20 Remaining Time: 10 Antireplay Sequence Number: 40
IKE status: Established Counters: Ann msgs sent: 31 Ann msgs sent with reply request: 2 Ann msgs recv: 64 Ann msgs recv with reply request: 1 Packet sent drops: 7 Packet Recv drops: 0 Total bytes sent: 20887 Total bytes recv: 40244
In a functional COOP setup, this protocol flow should be observed:
IKE Exchange > ANN with COOP priorities exchanged > COOP Election > ANN from primary to secondary KS (policy, GM database, and keys)
When COOP does not work correctly, or if there is a COOP split, such as multiple KSs become the primary KS, these debugs must be collected for troubleshooting:
debug crypto isakmp debug crypto gdoi ks coop all-levels show crypto isakmp sa show crypto gdoi ks coop
Successful IKE exchange is required for GETVPN in order to secure the control channel for the subsequent policy and SA download. At the end of the successful IKE exchange, a GDOI_REKEY sa is created.
In versions earlier than Cisco IOS 15.4(1)T, the GDOI_REKEY can be shown with the show crypto isakmp sa command:
GM1#show crypto isakmp sa IPv4 Crypto ISAKMP SA dst src state conn-id status 10.1.13.2 10.1.11.2 GDOI_REKEY 1075 ACTIVE 10.1.11.2 10.1.13.2 GDOI_IDLE 1074 ACTIVE
IPv6 Crypto ISAKMP SA
In Cisco IOS 15.4(1)T and later, this GDOI_REKEY sa is shown with the show crypto gdoi rekey sa command:
GM1#show crypto gdoi rekey sa
GETVPN REKEY SA
dst src conn-id status
10.1.13.2 10.1.11.2 1114 ACTIVE
Note: Once the initial IKE exchange completes, subsequent policies and keys will be pushed from the KS to the GM with the use of the GDOI_REKEY SA. So there is no rekey for the GDOI_IDLE SA when they expire; they disappear when their lifetimes expire. However, there should always be GDOI_REKEY SA on the GM in order for it to receive rekeys.
The IKE exchange for GETVPN is no different from the IKE used in traditional point-to-point IPsec tunnels, so the troubleshooting method remains the same. These debugs must be collected in order to troubleshoot IKE authentication issues:
debug crypto isakmp debug crypto isakmp error debug crypto isakmp detail (hidden command, if detailed isakmp exchange information is needed) debug crypto isakmp packet (hidden command, if packet level isakmp information is needed)
Registration, Policy Download, and SA Install
Once IKE authentication succeeds, GM registers with the KS. These syslog messages are expected to be seen when this occurs correctly:
%GDOI-5-GM_REKEY_TRANS_2_UNI: Group G1 transitioned to Unicast Rekey. %GDOI-5-SA_KEK_UPDATED: SA KEK was updated %GDOI-5-SA_TEK_UPDATED: SA TEK was updated %GDOI-5-GM_REGS_COMPL: Registration to KS 10.1.12.2 complete for group G1 using address 10.1.13.2 %GDOI-5-GM_INSTALL_POLICIES_SUCCESS: SUCCESS: Installation of Reg/Rekey policies from KS 10.1.12.2 for group G1 & gm identity 10.1.13.2
The policy and keys can be verified with this command:
GM1#show crypto gdoi GROUP INFORMATION
Group Name : G1 Group Identity : 3333 Crypto Path : ipv4 Key Management Path : ipv4 Rekeys received : 1 IPSec SA Direction : Both
Group Server list : 10.1.11.2 10.1.12.2
Group member : 10.1.13.2 vrf: None Version : 1.0.4 Registration status : Registered Registered with : 10.1.12.2 Re-registers in : 139 sec Succeeded registration: 1 Attempted registration: 1 Last rekey from : 10.1.11.2 Last rekey seq num : 0 Unicast rekey received: 1 Rekey ACKs sent : 1 Rekey Rcvd(hh:mm:ss) : 00:05:20 allowable rekey cipher: any allowable rekey hash : any allowable transformtag: any ESP
Rekeys cumulative Total received : 1 After latest register : 1 Rekey Acks sents : 1
ACL Downloaded From KS 10.1.11.2: access-list deny icmp any any access-list deny eigrp any any access-list deny ip any 126.96.36.199 0.255.255.255 access-list deny ip 188.8.131.52 0.255.255.255 any access-list deny udp any port = 848 any port = 848 access-list permit ip any any
KEK POLICY: Rekey Transport Type : Unicast Lifetime (secs) : 878 Encrypt Algorithm : 3DES Key Size : 192 Sig Hash Algorithm : HMAC_AUTH_SHA Sig Key Length (bits) : 1024
TEK POLICY for the current KS-Policy ACEs Downloaded: Serial1/0: IPsec SA: spi: 0x8BF147EF(2347845615) transform: esp-3des esp-sha-hmac sa timing:remaining key lifetime (sec): (200) Anti-Replay(Time Based) : 4 sec interval
GM1# GM1# GM1#show crypto ipsec sa
interface: Serial1/0 Crypto map tag: gm1map, local addr 10.1.13.2
Once the GMs are registered to the KS and the GETVPN network is properly set up, the primary KS is responsible for sending rekey messages to all the GMs registered to it. The rekey messages are used in order to synchronize all the policies, keys, and pseudotimes on the GMs. The rekey messages can be sent through a unicast or a multicast method.
This syslog message is seen on the KS when the rekey message is sent:
%GDOI-5-KS_SEND_UNICAST_REKEY: Sending Unicast Rekey for group G1 from address 10.1.11.2 with seq # 11
On the GMs, this is the syslog that is seen when it receives the rekey:
%GDOI-5-GM_RECV_REKEY: Received Rekey for group G1 from 10.1.11.2 to 10.1.20.2 with seq # 11
RSA Key Pair Requirement for Rekey on KS
Rekey functionality requires the presence of RSA keys on the KS. The KS provides the public key of the RSA key pair to the GM through this secure channel during registration. The KS then signs the GDOI messages sent to the GM with the private RSA key in the GDOI SIG payload. The GM receives the GDOI messages and uses the public RSA key in order to verify the message. The messages between the KS and the GM are encrypted with the KEK, which is also distributed to the GM during registration. Once the registration is complete, subsequent rekeys are encrypted with the KEK and signed with the private RSA key.
If the RSA key is no present on the KS during GM registration, this message appears on the syslog:
%GDOI-1-KS_NO_RSA_KEYS: RSA Key - get : Not found, Required for group G1
When the keys are not present on the KS, the GM registers for the first time, but the next rekey fails from the KS. Eventually the existing keys on the GM expire, and it reregisters again.
%GDOI-4-GM_RE_REGISTER: The IPSec SA created for group G1 may have expired/been cleared, or didn't go through. Re-register to KS.
Since the RSA key pair is used in order to sign the rekey messages, they MUST be the same between the primary and all secondary KSs. This ensures that during a primary KS failure, the rekeys sent by a secondary KS (the new primary KS) can still be properly validated by the GMs. When it generates the RSA key pair on the primary KS, the key pair must be created with the exportable option so that they can be exported to all the secondary KSs in order to meet this requirement.
KEK/TEK rekey failure is one of the most common GETVPN problems encountered in customer deployments. Troubleshooting rekey issues should follow the rekey steps as outlined here:
Did the rekeys get sent by the KS?
This can be checked by an observion of the %GDOI-5-KS_SEND_UNICAST_REKEY syslog message or more accurately with this command:
KS1#show crypto gdoi ks rekey Group G1 (Unicast) Number of Rekeys sent : 341 Number of Rekeys retransmitted : 0 KEK rekey lifetime (sec) : 1200 Remaining lifetime (sec) : 894 Retransmit period : 10 Number of retransmissions : 5 IPSec SA 1 lifetime (sec) : 900 Remaining lifetime (sec) : 405
The number of rekeys retransmitted is indicative of rekey acknowledgment packets not received by the KS and therefore possible rekey issues. Keep in mind that the GDOI rekey uses UDP as an unreliable transport mechanism, so some rekey drops might be expected depending on the reliability of the underlying transport network, but a trend of increasing rekey retransmissions should always be investigated.
More detailed per-GM rekey statistics can also be obtained. This is typically the first place to look for potential rekey issues.
KS1#show crypto gdoi ks members
Group Member Information :
Number of rekeys sent for group G1 : 346
Group Member ID : 10.1.14.2 GM Version: 1.0.4 Group ID : 3333 Group Name : G1 Key Server ID : 10.1.11.2 Rekeys sent : 346 Rekeys retries : 0 Rekey Acks Rcvd : 346 Rekey Acks missed : 0
Sent seq num : 2 1 2 1 Rcvd seq num : 2 1 2 1
Group Member ID : 10.1.13.2 GM Version: 1.0.4 Group ID : 3333 Group Name : G1 Key Server ID : 10.1.12.2 Rekeys sent : 340 Rekeys retries : 0 Rekey Acks Rcvd : 340 Rekey Acks missed : 0
Sent seq num : 2 1 2 1 Rcvd seq num : 2 1 2 1
Did the rekey packets get delivered in the underlying infrastructure network?
Standard IP troubleshooting along the rekey forwarding path should be followed in order to ensure the rekey packets are not dropped in the transit network between KS and GM. Some common troubleshooting tools used here are input/output Access Control Lists (ACLs), Netflow, and packet capture in the transit network.
Did the rekey packets reach the GDOI process for rekey processing?
Check the GM rekey statistics:
GM1#show crypto gdoi gm rekey Group G1 (Unicast) Number of Rekeys received (cumulative) : 340 Number of Rekeys received after registration : 340 Number of Rekey Acks sent : 340
Did the rekey acknowledgement packet return to the KS?
Follow Steps 1 through 3 in order to trace the rekey acknowledgement packet from the GM back to the KS.
Multicast rekey is different from unicast rekey in these aspects:
Since multicast is used in order to transport these rekey packets from the KS to the GMs, the KS does not need to replicate the rekey packets itself. The KS only sends one copy of the rekey packet, and they are replicated in the multicast-enabled network.
There is no acknowledgement mechanism for multicast rekey, so if a GM were not to receive the rekey packet, the KS would have no knowledge of it, and therefore will never remove a GM from its GM database. And because there is no acknowledgement, the KS will always retransmit the rekey packets based on its rekey retransmission configuration.
The most commonly seen multicast rekey problem is when the rekey is not received on the GM. There could be a number of possible causes for this, such as:
Packet delivery issue within the multicast routing infrastructure
End-to-end multicast routing is not enabled within the network
The first step to troubleshoot an issue with multicast rekey is to see if rekey works when switched from the multicast to the unicast method.
Once you identify that the issue is specific to multicast rekey, verify that KS sends the rekey to the multicast address specified.
%GDOI-5-KS_SEND_MCAST_REKEY: Sending Multicast Rekey for group G1 from address 10.1.11.2 to 184.108.40.206 with seq # 6
Test multicast connectivity between the KS and GM with an Internet Control Message Protocol (ICMP) request to the multicast address. All the GMs that are part of the multicast group should reply to the ping. Ensure that ICMP is excluded from the KS encryption policy for this test.
Type escape sequence to abort. Sending 1, 100-byte ICMP Echos to 220.127.116.11, timeout is 2 seconds:
Reply to request 0 from 10.1.21.2, 44 ms
If the multicast ping test fails, then multicast troubleshooting must be performed, which is outside of the scope of this document.
Control Plane Relay Check
When customers upgrade their GM to a new Cisco IOS version, they might experience KEK rekey failures with this message observed in the syslog:
%GDOI-3-GDOI_REKEY_SEQ_FAILURE: Failed to process rekey seq # 1 in seq payload for group G1, last seq # 11 %GDOI-3-GDOI_REKEY_FAILURE: Processing of REKEY payloads failed on GM 10.1.13.2 in the group G1, with peer at 10.1.11.2 %CRYPTO-6-IKMP_MODE_FAILURE: Processing of GDOI mode failed with peer at 10.1.11.2
This behavior is caused by an interoperability issue introduced with the anti-replay check that is added for control plane messages. Specifically, a KS that runs the older code will reset the KEK rekey sequence number to 1, and this will be dropped by the GM that runs the new code when it interprets that as a replayed rekey packet. For more details, see Cisco bug ID CSCta05809 (GETVPN: GETVPN control-plane sensible to replay), and GETVPN Configuration Restrictions.
With GETVPN, the Control Plane messages can carry time-sensitive information in order to provide the time-based anti-replay check service. Therefore, these messages require anti-replay protection themselves in order to ensure time accruracy. These messages are:
Rekey Messages from KS to GM
COOP Announcement Messages between KSs
As part of this anti-replay protection implementation, sequence number checks were added in order to protect replayed messages, as well as a pseudotime check when TBAR is enabled.
In order to resolve this issue, both the GM and KS must be upgraded to Cisco IOS versions after the Control Plane replay check feature. With the new Cisco IOS code, KS does not reset the sequence number back to 1 for a KEK rekey, but instead it continues to use the current sequence number and only resets the sequence number for TEK rekeys.
These Cisco IOS versions have the Replay Check features:
15.0(1)M and later
Other Replay Related Issues
COOP failure due to ANN messages failing replay check (Cisco bug ID CSCtc52655)
Debug Control Plane Replay Failures
For other Control Plane Replay failures, collect this information and make sure the times are synched between the KS and GM.
Syslog from both GM and KS
GDOI debugs (rekey and replay) from both KS and GM
Control Plane Packet Fragmentation Issues
With GETVPN, Control Plane Packet fragmentation is a common issue, and it can manifest itself in one of these two scenarios when the Control Plane packets are large enough that they will require IP fragmentation:
GETVPN COOP Announcement packets
GETVPN rekey packets
COOP Announcement Packets
The COOP Announcement packets carry the GM database information, and thus can grow big in a large GETVPN deployment. From past experience, a GETVPN network that consists of 1500+ GMs will produce Announcement packets larger than 18024 bytes, which is the Cisco IOS default Huge buffer size. When this happens, the KS fails to allocate a buffer large enough to transmit the ANN packets with this error:
In order to rectify this condition, this buffer tuning is recommended:
buffers huge permanent 10 buffers huge size 65535
GETVPN rekey packets can also exceed the typical 1500 IP Maximum Transition Unit (MTU) size when the encryption policy is large, such as a policy that consists of 8+ lines of Access Control Entries (ACEs) in the encryption ACL.
Fragmentation Problem and Identification
In both of the previous scenarios, GETVPN must be able to properly transmit and receive the fragmented UDP packets in order for COOP or GDOI rekey to work properly. IP fragmentation can be a problem in some network environments. For example, a network that consists of Equal Cost Multi Path (ECMP) forwarding plane, and some devices in the forwarding plane require virtual reassembly of the fragmented IP packets, such as Virtual Fragmentation Reassembly (VFR).
In order to identify the problem, check the reassembly errors on the device where it is suspected that the fragmented UDP 848 packets are not properly received:
If the reassembly timeouts continue to increment, use the debug ip error command in order to confirm if the drop is part of the rekey/COOP packet flow. Once confirmed, normal IP forwarding troubleshooting should be performed in order to isolate the exact device in the forwarding plane that might have dropped the packets. Some commonly used tools include:
Traffic forwarding statistics
Security feature statistics (Firewall, IPS)
GDOI Interoperability Issues
Various interoperability issues have been found with GETVPN over the years, and it is critical to notice the Cisco IOS release versions between KS and GM and amongst the KSs for interoperability issues.
Other well known GETVPN interoperability issues are:
Cisco bug ID CSCub42920 (GETVPN: KS fails to validate hash in rekey ACK from previous GM versions)
Cisco bug ID CSCuw48400 (GetVPN GM unable to register or rekey fails - sig-hash > default SHA-1)
Cisco bug ID CSCvg19281 ( Multiple GETVPN GM crashes after migration to new KS pair ; if a GM version is earlier than 3.16, and KS is upgraded from an earlier code to 3.16 or later, this issue can happen)
GETVPN IOS Upgrade Procedure
This Cisco IOS upgrade procedure should be followed when a Cisco IOS code upgrade needs to be performed in a GETVPN environment:
Upgrade a secondary KS first and wait until COOP KS election is completed.
Repeat Step1 for all secondary KSs.
Upgrade the primary KS.
Troubleshoot GETVPN Data Plane Issues
Compared to Control Plane problems, GETVPN data plane issues are problems where the GM has the policy and keys to perform dataplane encryption and decryption, but for some reason the end-to-end traffic flow does not work. Most of the dataplane issues for GETVPN relate to generic IPsec forwarding, and are not GETVPN specific. So most of the troubleshooting approach described here applies to generic IPsec dataplane issues as well.
With encryption problems (both Group-based or pair-wise tunnels), it is important to troubleshoot the problem and isolate the problem to a particular part of the datapath. Specifically, the troubleshooting approach described here is intended to help you answer these questions:
Which device is the culprit - encrypting router or decrypting router?
In which direction is the problem happening - ingress or egress?
GETVPN Data Plane Troubleshooting Tools
IPsec dataplane troubleshooting is very different from that for the Control Plane. With the dataplane, there are usually no debugs that you can run, or at least run safely in a production environment. So the troubleshooting relies heavily on different counters and traffic statistics that can help trace the packet along a forwarding path. The idea is to be able to develop a set of checkpoints in order to help isolate where packets might be dropped as shown here:
Here are some data plane debugging tools:
IP Precedence Accounting
IP Cisco Express Forwarding (CEF) Global and Per-feature Drop Counters
Embedded Packet Capture (EPC)
Data Plane Debugs (IP packet and CEF debugs)
The checkpoints in the datapath in the previous image can be validated with these tools:
Crypto Engine show crypto ipsec sa show crypto ipsec sa detail show crypto engine accelerator statistics
Egress LAN interface Egress netflow Embedded packet capture
The return path follows the same traffic flow. The next sections have some examples of these dataplane tools in use.
The encryption/decryption counters on a router are based on an IPsec flow. Unfortunately this does not work well with GETVPN since GETVPN typically deploys a "permit ip any any" encryption policy that encrypts everything. So if the problem only happens for some of the flows and not all, these counters can be somewhat difficult to use in order to correctly assess if the packets are encrypted or decrypted when there is enough significant background traffic that works.
Netflow can be used in order to monitor both the ingress and egress traffic on both GMs. Note with the GETVPN permit ip any any policy, the enrypted traffic will be aggregate and does not provide the per-flow information. Per-flow information will then need to be collected with the DSCP/precedence marking described later.
In this example, the netflow for a 100 count ping from a host behind GM1 to a host behind GM2 is shown at the various checkpoints.
interface Ethernet0/0 description LAN ip address 192.168.13.1 255.255.255.0 ip flow ingress ip pim sparse-dense-mode ! interface Serial1/0 description WAN interface ip address 10.1.13.2 255.255.255.252 ip flow egress ip pim sparse-dense-mode crypto map gm1map
Note: In the previous output, * denotes egress traffic. The first line shows egress encrypted traffic (with protocol 0x32 = ESP) out of the WAN inteface, and the second line ingress ICMP traffic hitting the LAN interface.
interface Ethernet0/0 description LAN interface ip address 192.168.14.1 255.255.255.0 ip flow egress ip pim sparse-dense-mode ! interface Serial1/0 description WAN interface ip address 10.1.14.2 255.255.255.252 ip flow ingress ip pim sparse-dense-mode crypto map gm1map
The challenge with troubleshooting an encryption problem is that once the packet is encrypted you lose visibility into the payload, which is what encryption is supposed to do, and that makes it difficult to trace the packet for a particular IP flow. There are two ways to address this limitation when it comes to troubleshooting an IPsec problem:
Use ESP-NULL as the IPsec transform. IPsec still performs ESP encapsulation but no encryption is applied to the payload, so they are visible in a packet capture.
Mark an IP flow with a unique Differentiated Services Code Point (DSCP)/precedence marking based on their L3/L4 characteristics.
ESP-NULL require changes on both tunnel end points and often is not allowed based on the customer security policy. Therefore, Cisco typically recommends the use of DSCP/precedence marking instead.
DSCP/Precedence Reference Chart
7 Network Control
6 Internetwork Control
4 Flash Override
Mark Packets with DSCP/Precedence
These methods are typically used in order to mark packets with the specific DSCP/Precedence markings.
interface Ethernet1/0 ip policy route-map mark ! access-list 150 permit ip host 172.16.1.2 host 172.16.254.2 ! route-map mark permit 10 match ip address 150 set ip precedence flash-override
class-map match-all my_flow match access-group 150 ! policy-map marking class my_flow set ip precedence 4 ! interface Ethernet1/0 service-policy input marking
GM1-host#ping ip Target IP address: 192.168.14.2 Repeat count : Datagram size : Timeout in seconds : Extended commands [n]: y Source address or interface: Type of service : 136 ... <snip>
Note: It is always a good idea to monitor the normal traffic flow and DSCP/precedence profile before you apply marking so that the marked traffic flow is unique.
Monitor Marked Packets
IP Precedence Accounting
interface Ethernet0/0 ip address 192.168.1.2 255.255.255.0 ip accounting precedence input
middle_router#show access-list 144 Extended IP access list 144 10 permit ip any any precedence routine 20 permit ip any any precedence priority 30 permit ip any any precedence immediate 40 permit ip any any precedence flash 50 permit ip any any precedence flash-override (100 matches) 60 permit ip any any precedence critical 70 permit ip any any precedence internet (1 match) 80 permit ip any any precedence network
Embedded Packet Capture
Embedded Packet Capture (EPC) is a useful tool to capture packets at the interface level in order to identify if a packet has reached a specific device. Remember that EPC works well for clear text traffic, but it can be a challenge when the captured packets are encrypted. Therefore techniques like DSCP/precedence marking discussed previously or other IP characters, such as the length of the IP packet, have to be used together with EPC in order to make the troubleshooting more effective.
Cisco IOS-XE Packet Trace
This is a useful feature to trace the feature forwarding path on all platforms that run Cisco IOS-XE, such as CSR1000v, ASR1000, and ISR4451-X.
GETVPN Data Plane Common Issues
Troubleshooting the IPsec dataplane for GETVPN is mostly no different from troubleshooting traditional point-to-point IPsec dataplane issues, with two exceptions due to these unique dataplane properties of GETVPN.
Time Based Anti-Replay Failure
In a GETVPN network, TBAR failures can often be difficult to troubleshoot since there are no longer pair-wise tunnels. In order to troubleshoot GETVPN TBAR failures, complete these steps:
Identify which packet is dropped due to TBAR failure and subsequently identify the encrypting GM.
Prior to Version 15.3(2)T, the TBAR failure syslog did not print the source address of the failed packet, so this makes it very difficult to identify which packet failed. This has been significantly improved in Version 15.3(2)T and later, where Cisco IOS prints this:
%GDOI-4-TIMEBASED_REPLAY_FAILED: An anti replay check has failed in group G1: my_pseudotime = 620051.84 secs, peer_pseudotime = 619767.09 secs, replay_window = 4 (sec), src_ip = 192.168.13.2, dst_ip = 192.168.14.2
A TBAR history was also implemented in this version:
GM2#show crypto gdoi gm replay Anti-replay Information For Group G1: Timebased Replay: Replay Value : 621388.66 secs Input Packets : 0 Output Packets : 0 Input Error Packets : 2 Output Error Packets : 0 Time Sync Error : 0 Max time delta : 0.00 secs
TBAR Error History (sampled at 10pak/min): 19:29:32.081 EST Wed Nov 13 2013: src=192.168.13.2; my_pst=620051.84 secs; peer_pst=619767.09 secs; win=4
Note: The enhancements mentioned previously have since been implemented in Cisco IOS-XE by Cisco bug ID CSCun49335 and in Cisco IOS by Cisco bug ID CSCub91811.
For Cisco IOS versions that did not have this feature, debug crypto gdoi gm replay detail can also provide this information, although this debug prints the TBAR information for all traffic (not only packets dropped due to TBAR failure), so it might not be feasible to run in a production environment.
GDOI:GM REPLAY:DET:(0):my_pseudotime is 621602.30 (secs), peer_pseudotime is 621561.14 (secs), replay_window is 4 (secs), src_addr = 192.168.14.2, dest_addr = 192.168.13.2
Once the source of the packet is identified, you should be able to find the encrypting GM. Then, the pseudotimestamp on both the encrypting and decrypting GMs should be monitored for any potential pseudotime drift. The best way to do this would be to synchronize both GMs and the KS to NTP and periodically collect the pseudotime information with a reference system clock on all of them in order to determine if the problem is caused by clock skew on the GMs.
GM1#show crypto gdoi gm replay Load for five secs: 0%/0%; one minute: 0%; five minutes: 0% Time source is hardware calendar, *21:06:26.469 EST Wed Nov 13 2013
Anti-replay Information For Group G1: Timebased Replay: Replay Value : 625866.26 secs Input Packets : 0 Output Packets : 0 Input Error Packets : 0 Output Error Packets : 0 Time Sync Error : 0 Max time delta : 0.00 secs
GM2#show crypto gdoi gm replay Load for five secs: 0%/0%; one minute: 0%; five minutes: 0% Time source is hardware calendar, *21:06:26.743 EST Wed Nov 13 2013
Anti-replay Information For Group G1: Timebased Replay: Replay Value : 625866.51 secs Input Packets : 4 Output Packets : 4 Input Error Packets : 2 Output Error Packets : 0 Time Sync Error : 0 Max time delta : 0.00 secs
In the previous example, if the pseudotime (as indicated by Replay Value) is significantly different between the GMs when the outputs are captured with the same reference time, then the problem can be attributed to clock skew.
Note: On the Cisco Aggregated Services Router 1000 Series platform, due to the platform architecture, the datapath on the Quantum Flow Processor (QFP) actually refers to the wall clock for counting pseudotime ticks. This has created problems with TBAR when the wall clock time changes due to NTP sync. This problem is documented with Cisco bug ID CSCum37911.
PMTUD and GETVPN Header Preservation
With GETVPN, Path MTU Discovery (PMTUD) does not work between the encrypting and decrypting GMs, and large packets with the Don't Fragment (DF) bit set can get blackholed. The reason that this does not work is due to GETVPN Header Preservation where the data source/destination addresses are preserved in the ESP encapsulating header. This is depicted in this image:
As the image shows, PMTUD breaks down with GETVPN with this flow:
Large data packet arrives on the encrypting GM1.
The post-encryption ESP packet is forwarded out of GM1 and delivered towards the destination.
If there is a transit link with IP MTU of 1400 bytes, the ESP packet will be dropped, and an ICMP 3/4 packet too big message will be sent towards the packet source, which is the source of the data packet.
The ICMP3/4 packet is either dropped due to ICMP not excluded from the GETVPN encryption policy, or dropped by the end host since it does not know anything about the ESP packet (unauthenticated payload).
In summary, PMTUD does not work with GETVPN today. In order to work around this issue, Cisco recommends these steps:
Implement "ip tcp adjust-mss" in order to reduce the TCP packet segment size tin order o accommodate encryption overhead and minimum path MTU in the transit network.
Clear the DF bit in the data packet as they arrive on the encrypting GM in order to avoid PMTUD.
This message can be generated when an IPsec packet is received that does not match an SPI in the SADB. See Cisco bug ID CSCtd47420 - GETVPN - CRYPTO-4-RECVD_PKT_NOT_IPSEC reported for pkt not matching flow. An example is:
%CRYPTO-4-RECVD_PKT_NOT_IPSEC: Rec'd packet not an IPSEC packet. (ip) vrf/dest_addr= /192.168.14.2, src_addr= 192.168.13.2, prot= 50
This message should be %CRYPTO-4-RECVD_PKT_INV_SPI, which is what gets reported for traditional IPsec as well as on some hardware platforms such as ASR. This cosmetic issue was fixed by Cisco bug ID CSCup80547: Error in reporting CRYPTO-4-RECVD_PKT_NOT_IPSEC for ESP pak.
Note: These messages can sometimes appear due to another GETVPN bug CSCup34371: GETVPN GM stops decrytping traffic after TEK rekey.
In this case, the GM cannot decrypt GETVPN traffic, although it has a valid IPsec SA in the SADB (the SA being rekeyed). The problem disappears as soon as the SA expires and is removed from the SADB. This issue causes significant outage, because TEK rekey is performed in advance. For example, the outage can be 22 minutes in the case of a TEK lifetime of 7200 seconds. See the bug description for the exact condition that should be met in order to encounter this bug.
Troubleshoot GETVPN on Platforms that Run Cisco IOS-XE
Platforms that run Cisco IOS-XE have platform-specific implementations, and often require platform-specific debugging for GETVPN issues. Here are a list of commands typically used in order to troubleshoot GETVPN on these platforms:
show crypto eli all
show platform software ipsec policy statistics
show platform software ipsec fp active inventory
show platform hardware qfp active feature ipsec spd all
show platform hardware qfp active statistics drop clear
show platform hardware qfp active feature ipsec data drop clear
An ASR1000 GM might continue to register to the Key Server if the crypto engine does not support the IPsec policy or algorithm received. For example, on Nitrox based ASR platforms (such as ASR1002), Suite-B or SHA2 policies are not supported and this can cause the continuous re-registration symptoms.
This enhancement bug has been opened to lift this restriction, Cisco bug ID CSCuq25476 - ASR1k needs to support a GETVPN TBAR window size of less than 20 seconds.
Update: This restriction has since been lifted with the fix for Cisco bug ID CSCur57558 , and it is no longer a limitation in XE3.10.5, XE3.13.2 and later code.
Also note, for a GM that runs on Cisco IOS-XE platforms (ASR1k or ISR4k), it is highly recommended that the device runs a version with the fix for this issue if TBAR is enabled; Cisco bug ID CSCut91647 - GETVPN on IOS-XE: GM incorrectly drops packets due to TBAR failure.
ISR4x00 Classification Issue
A regression was found on the ISR4x00 platform where the deny policies are ignored. For details, see Cisco bug ID CSCut14355 - GETVPN - ISR4300 GM ignores deny policy.