Strange index error on NFS after upgrade to CentOS 7 from 6

Alessio Cecchi alessio at skye.it
Tue Aug 25 17:56:06 EEST 2020


Il 25/08/20 16:27, Aki Tuomi ha scritto:
>> On 25/08/2020 16:43 Alessio Cecchi <alessio at skye.it> wrote:
>>
>>
>> Hi,
>>   
>>   I'm running 3 Dovecot director and 6 Dovecot backend on CentOS 6 and Dovecot 2.3.10, with Maildir shared on NFSv3 (on NetApp).
>>   
>>   Since CentOS 6 will be EOL on November I started to upgrade to CentOS 7 the first Dovecot backend.
>>   
>>   But after Director start to direct users on new CentOS 7 server some errors, for any users, come up, an example:
>>   
>>   Aug 21 14:56:41 Error: imap(alessio at cecchi.it) session=<11sBlmKtj1qelGW/>: Mailbox INBOX: Broken or unexpectedly changed file /home/vmail/domains/cecchi.it/alessio/Maildir/dovecot-uidlist line 18664: Invalid data: - re-reading from beginning
>>   Aug 21 14:56:41 Error: imap(alessio at cecchi.it) session=<11sBlmKtj1qelGW/>: Mailbox INBOX: Broken file /home/vmail/domains/cecchi.it/alessio/Maildir/dovecot-uidlist line 18664: Invalid data:
>>   
>>   Consequently to the error the user see the synchronization of his mailbox again.
>>   
>>   The Dovecot version and configuration is the same that in CentOS 6, also the mount parameters are the same, so I suspect that could be some difference in the NFS client implementation on CentOS 7.
>>   
>>   This the entry on fstab:
>>   
>>   192.168.1.2:/vmail0 /mnt/vmail0 nfs rw,nfsvers=3,noatime,nodiratime,_netdev,nordirplus 0 0
>>   # Bind for vmail0
>>   /mnt/vmail0/domains /home/vmail/domains none x-systemd.requires=/mnt/vmail0,x-systemd.automount,bind 0 0
>>   
>>   In the archive of this list I found a similar issue but without a solutions:
>>   
>>   https://dovecot.org/pipermail/dovecot/2018-October/113207.html
>>   
>>   Do you have any suggestions?
>>   Thanks
>>
>> -- 
>> Alessio Cecchi
>> Postmaster @ http://www.qboxmail.it
>> https://www.linkedin.com/in/alessice
> What versions are you running mixed?
>
> Aki

Hi Aki,

the only mixed version is CentOS version on Dovecot backend:

CentoS 6 backend:

# dovecot -n
# 2.3.10 (0da0eff44): /etc/dovecot/dovecot.conf
# Pigeonhole version 0.5.10 (bf8ef1c2)
# OS: Linux 2.6.32-754.29.2.el6.x86_64 x86_64 CentOS release 6.10 (Final)

Centos 7 backend:

# dovecot -n
# 2.3.10 (0da0eff44): /etc/dovecot/dovecot.conf
# Pigeonhole version 0.5.10 (bf8ef1c2)
# OS: Linux 3.10.0-1127.18.2.el7.x86_64 x86_64 CentOS Linux release 
7.8.2003 (Core)

and also the dovecot configuration is the same (was copied via rsync):

$ diff dovecot-a-cos6.txt dovecot-a-cos7.txt
3,4c3,4
< # OS: Linux 2.6.32-754.29.2.el6.x86_64 x86_64 CentOS release 6.10 (Final)
< # Hostname: pop01
---
 > # OS: Linux 3.10.0-1127.18.2.el7.x86_64 x86_64 CentOS Linux release 
7.8.2003 (Core)
 > # Hostname: pop02
126c126
< import_environment = TZ CORE_OUTOFMEM CORE_ERROR
---
 > import_environment = TZ CORE_OUTOFMEM CORE_ERROR LISTEN_PID LISTEN_FDS

I have noticed this, a user have the "Indalid data" error on Centos 7, I 
moved him to Centos 6, with doveadm director move, and the "Invalid 
data" error come again also on Centos 6 but only one time, like if the 
dovecot-uidlist file was wrong and need to be fixed, but after the "fix" 
executed from Dovecot in Centos 6 the file become fine.

Instead, on Centos 7 the "Invalid data" error comes multiple time.

Thanks

-- 
Alessio Cecchi
Postmaster @ http://www.qboxmail.it
https://www.linkedin.com/in/alessice

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://dovecot.org/pipermail/dovecot/attachments/20200825/b812e623/attachment.html>


More information about the dovecot mailing list