Avamar: Backups may stop responding, or file system backups fail with "Fatal Server Error occurred"

Backups are failing with the below logs being generated.

Avamar GSAN err.log may show:

[srvm-518654#srv:481]  WARN: <1479> sessionkeytable::setpublickey new serial=<x> less than current minserial=<y>
...
[srvm-525136#srv:775]  WARN: <1426> sessionkeytable::comparesignature incorrect signature

Avtar logs may show:

2019-12-31 07:35:23 avtar Error <8609>: Connection killed from GSAN. (Log #1) 
2019-12-31 07:35:23 avtar FATAL <5704>: Fatal Server Error occurred (MSG_ERR_AUTH_FAIL), aborting execution (SECURETICKETLOGIN=452 serial=1 seq=0 flags=R:H:0 kind=0 rsp=MSG_ERR_AUTH_FAIL) (Log #1) 
2019-12-31 07:35:23 avtar Error <5126>: Login error 5: Authorization failure (Session Ticket login) (Log #1) 
2019-12-31 07:35:23 avtar FATAL <8941>: Fatal server connection problem, aborting initialization. Verify correct server address and login credentials. (Log #1) 
2019-12-31 07:35:23 avtar Error <7001>: Exiting avtar with run-at-end script failure -1 (Log #1)

9-12-31 07:35:23 avtar Info <8474>: - Log file path: /usr/local/avamar/var/clientlogs/some-Unix.log
2019-12-31 07:35:23 avtar Info <6555>: Initializing connection
2019-12-31 07:35:23 avtar Info <5552>: Connecting to Avamar Server (avamar.com)
2019-12-31 07:35:23 avtar Info <5554>: Connecting to one node in each datacenter
2019-12-31 07:35:23 avtar Info <5993>: - Connect: Connected to 10.x.x.x:29000, Priv=0, SSL Cipher=AES256-SHA
2019-12-31 07:35:23 avtar Info <5993>: - Datacenter 0 has 1 nodes: Connected to 10.x.x.x:29000, Priv=0, SSL Cipher=AES256-SHA
2019-12-31 07:35:23 avtar Info <5581>: Logging in on connection 0 with Session Ticket
2019-12-31 07:35:23 avtar Info <18854>: Using Secure Session Ticket Format
2019-12-31 07:35:23 avtar Error <8609>: Connection killed from GSAN.
2019-12-31 07:35:23 avtar Info <9772>: Starting graceful (staged) termination, KILL event received (wrap-up stage)
2019-12-31 07:35:23 avtar FATAL <5704>: Fatal Server Error occurred (MSG_ERR_AUTH_FAIL), aborting execution (SECURETICKETLOGIN=452 serial=1 seq=0 flags=R:H:0 kind=0 rsp=MSG_ERR_AUTH_FAIL)
2019-12-31 07:35:23 avtar Error <5126>: Login error 5: Authorization failure (Session Ticket login)
2019-12-31 07:35:23 avtar FATAL <8941>: Fatal server connection problem, aborting initialization. Verify correct server address and login credentials.
2019-12-31 07:35:23 avtar Info <6149>: Error summary: 4 errors: 8941, 5704, 5126, 8609
2019-12-31 07:35:23 avtar Info <5917>: Back from run-at-end, exit code -1
2019-12-31 07:35:23 avtar Error <7001>: Exiting avtar with run-at-end script failure -1

You can see in the GSAN logs that the session ticket assigned may have already been used for a previous backup.
This could occur from a GSAN or MCS rollback putting the session ticket out of order.

From the GSAN log

WARN: <1418> sessionkeytable::setpublickey session ticket with serial=28451 was used already
WARN: <1426> sessionkeytable::comparesignature incorrect signature

When MCS has a larger session ticket than GSAN, we could infer that GSAN is behind and using old session tickets to give to clients.

Use the following grep command to help diagnose.

grep -i "sessionkeytable\|minserial" /data01/cur/gsan.log.00?

We must get the session tickets to sync up again.

Step 1:
Get the current GSAN minserial.

avmaint cat /sysinfo/security/keytable

Example:

root@avamar:/home/admin/#: avmaint cat /sysinfo/security/keytable
<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
<keytable minimalserial="<minserial_less_than_mcs>">

Step 2:
Get the current session ticket serial from MCS.

cat /usr/local/avamar/var/mc/server_data/prefs/sessionticket.srl

Example:

root@avamar:/home/admin/#: cat /usr/local/avamar/var/mc/server_data/prefs/sessionticket.srl
<mcs_session_ticket_larger_than_gsan>

Step 3:
Create a public key file with "mcecroot," and use it with avmaint.

cd /usr/local/avamar/lib

keytool -list -rfc -keystore /usr/local/avamar/lib/avamar_keystore -storepass `avlockbox.sh -r keystore_passphrase` -alias mcecroot | openssl x509 -pubkey -noout > mcecroot.pub

Example:
The command uses keytool to list the mcecroot alias in the avamar_keystore, piping the result to openssl to grab the public key from the key pair, and writing the public key to a file called mcecroot.pub.

root@avamar:/usr/local/avamar/lib/#: keytool -list -rfc -keystore /usr/local/avamar/lib/avamar_keystore -storepass `avlockbox.sh -r keystore_passphrase` -alias mcecroot | openssl x509 -pubkey -noout > mcecroot.pub

root@avamar:/usr/local/avamar/lib/#: cat mcecroot.pub
-----BEGIN PUBLIC KEY-----
MHYwEAYHKoZIzj0CAQYFK4EEACIDYgAEicGAqiHQQ8DRi0ZkgsvEA8fCEeqW/DIr
k53CFakwbOtTejb2Okk+2VTkY5H4gfLEKd4Gtq5gPP2OcpaUf/SrIc4MO3bn8OhC
l/vQKfAeJvrPPEvcIG0GiLyLtwJykeUq
-----END PUBLIC KEY-----

Step 4:
Write the new GSAN minserial using the mcecroot public key, suggest using a number larger than the current value in MCS.

avmaint publickey --keyfile=./mcecroot.pub --serial=<your number> --ava

Example:
We must choose a serial number high enough to avoid the following error.

root@avamar:/usr/local/avamar/lib/#: avmaint publickey --keyfile=./mcecroot.pub --serial=28500 --ava
ERROR: avmaint: publickey: server_exception(MSG_ERR_INVALID_PARAMETERS)

root@avamar:/usr/local/avamar/lib/#: avmaint publickey --keyfile=./mcecroot.pub --serial=38000 --ava

If you receive the invalid parameters error, keep choosing a higher number until there is no error.

Step 5:
Verify the change.

avmaint cat /sysinfo/security/keytable

Example:

root@avamar:/usr/local/avamar/lib/#: avmaint cat /sysinfo/security/keytable
<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
<keytable minimalserial="38000">     <-- new gsan minserial
[..]

Step 6:
Write that minserial to the MCS session ticket.

echo -n "<number you were able to use for mcec key>" > /usr/local/avamar/var/mc/server_data/prefs/sessionticket.srl

Example:

root@avamar:/usr/local/avamar/lib/#: echo -n "38000" > /usr/local/avamar/var/mc/server_data/prefs/sessionticket.srl

root@avamar:/usr/local/avamar/lib/#: cat /usr/local/avamar/var/mc/server_data/prefs/sessionticket.srl
38000

Step 7:
Restart MCS.

mcserver.sh --restart

Step 8:
Test backups.