Re: hostname in aliases.db
On 05/08/17 21:30, /dev/rob0 wrote: On Sat, Aug 05, 2017 at 07:58:19PM +0300, Marat Khalili wrote: That's what I'd like to know to, is this hostname mention even being used? I doubt it is, but I am too lazy / busy to test. :) You could also consult your Berkeley DB documentation. I do know that Postfix simply queries it for the localpart in a localpart@domain, where domain is in $mydestination. Metadata in aliases.db is not queried. Well, I looked it bit more into it and it quickly became exercise in software archaeology. That's what I found out: 1) It's not metadata, it's data associated with key 'YP_MASTER_NAME'. 2) It was present in sendmail and used by NIS: https://books.google.ru/books?id=NQblqMiVqvQC=PT152=PT152=YP_MASTER_NAME . 3) It was added to postfix in 1999 as documented in HISTORY: 19990325 Workaround: Solaris NIS alias maps need special entries (YP_MASTER_NAME, YP_LAST_MODIFIED). What's worse, normal keys/values include a null byte at the end, but the YP_XXX ones don't. Problem reported by Walcir Fontanini, state university of Campinas, Brazil. File: postalias/postalias.c. 4) Finally, it is currently set in postalias.c but never used indeed. Final results: no need to change it, but if necessary it can be changed with newaliases under chroot (tested this), directly with some Berkeley DB tool (since it is just one known key-value pair), or specified in makedbm command-line. Also, this problem is not even new: see https://www.ibm.com/support/knowledgecenter/en/ssw_aix_72/com.ibm.aix.nis/nis_movmastserv.htm Thank you for the help, it was an interesting excursion for me. -- With Best Regards, Marat Khalili
Re: hostname in aliases.db
On Sat, Aug 05, 2017 at 07:58:19PM +0300, Marat Khalili wrote: > > See also postalias(1), but I'm still not sure that this is a > > real problem. Does something in the container not work > > properly with host-generated aliases.db? > > That's what I'd like to know to, is this hostname mention even > being used? I doubt it is, but I am too lazy / busy to test. :) You could also consult your Berkeley DB documentation. I do know that Postfix simply queries it for the localpart in a localpart@domain, where domain is in $mydestination. Metadata in aliases.db is not queried. > Testing one particular container is not sufficient since I might > run into problems with some other container later, after I end > scripting it. > > > > The better way would probably be to simplify your mail > > infrastructure, using null clients where appropriate. > > > > I have nothing against containerizing Postfix nor running it > > in virtual machines, but unless your organization is very huge > > you do not need more than 1-2 MX hosts and perhaps a per-site > > MSA (which often can coexist on the submission port with MX > > instances.) > > Completely agree. It is mostly a problem of having a hammer and > seeing everything as a nail: I'm also not happy about having many > full-blown postfix instances, but it works and learning something > requires an effort. Hehe, okay. :) > Is msmtp the recommended tool for doing this or just one of the > many out there? There are several, and I am unable specifically to recommend one against the others, because I'm like you. I have this hammer, and when I need to do something involving sending mail, I just use Postfix. ;) -- http://rob0.nodns4.us/ Offlist GMX mail is seen only if "/dev/rob0" is in the Subject:
Re: hostname in aliases.db
You might indeed want to generate your aliases.db for each container, and chroot(1) might be a means to do that. Completely forgot about chroot, much better than starting whole container. See also postalias(1), but I'm still not sure that this is a real problem. Does something in the container not work properly with host-generated aliases.db? That's what I'd like to know to, is this hostname mention even being used? Testing one particular container is not sufficient since I might run into problems with some other container later, after I end scripting it. The better way would probably be to simplify your mail infrastructure, using null clients where appropriate. I have nothing against containerizing Postfix nor running it in virtual machines, but unless your organization is very huge you do not need more than 1-2 MX hosts and perhaps a per-site MSA (which often can coexist on the submission port with MX instances.) Completely agree. It is mostly a problem of having a hammer and seeing everything as a nail: I'm also not happy about having many full-blown postfix instances, but it works and learning something requires an effort. Is msmtp the recommended tool for doing this or just one of the many out there? -- With Best Regards, Marat Khalili
Re: hostname in aliases.db
On Sat, Aug 05, 2017 at 07:11:08PM +0300, Marat Khalili wrote: > I'm cloning an LXC container which optionally can contain postfix > installation. After cloning the filesystem there's a number of > places I need to change the hostname in. > > I used grep to search for these places and unexpectedly found > mentioning of hostname in /etc/aliases.db, even though /etc/aliases > does not include it. Is this an actual problem? Also, I wonder why you'd need multiple containers with Postfix installs? Did you consider possibly using a null client like msmtp, if all these containers need to do is send mail through a relayhost? > Thus I wonder if I need to re-generate /etc/aliases.db and how can > I do it without actually starting container? You might indeed want to generate your aliases.db for each container, and chroot(1) might be a means to do that. > I can run `newaliases -oAhash:/container/rootfs/etc/aliases` from > host, but then there's a name of the host system in aliases.db, > not container's. See also postalias(1), but I'm still not sure that this is a real problem. Does something in the container not work properly with host-generated aliases.db? > I can also re-generate it from within a container after starting > it and then reload postfix, but it is kludgy. Is there some better > way? The better way would probably be to simplify your mail infrastructure, using null clients where appropriate. I have nothing against containerizing Postfix nor running it in virtual machines, but unless your organization is very huge you do not need more than 1-2 MX hosts and perhaps a per-site MSA (which often can coexist on the submission port with MX instances.) -- http://rob0.nodns4.us/ Offlist GMX mail is seen only if "/dev/rob0" is in the Subject:
hostname in aliases.db
I've got the following problem which is not critical but still interesting. I'm cloning an LXC container which optionally can contain postfix installation. After cloning the filesystem there's a number of places I need to change the hostname in. I used grep to search for these places and unexpectedly found mentioning of hostname in /etc/aliases.db, even though /etc/aliases does not include it. Thus I wonder if I need to re-generate /etc/aliases.db and how can I do it without actually starting container? I can run `newaliases -oAhash:/container/rootfs/etc/aliases` from host, but then there's a name of the host system in aliases.db, not container's. I can also re-generate it from within a container after starting it and then reload postfix, but it is kludgy. Is there some better way? -- With Best Regards, Marat Khalili
