George Shuklin's questions

I want to create a reverse proxy under one location for two different upstreams. Upstreams are so different that they have different prefixes in URL, but the rest is the same.

• https://foo.example.com/foo/api/get/this
• https://bar.example.net/bar/api/get/this

Non-working configuration:

    upstream foo{
        server https://foo.example.com/foo;
        server https://bar.example.net/bar backup;
    }

    server{
      ...
      location /foobar/(.+) {
         proxy_pass https://foo/$1?args
      }
   }

I know I can't use URL in 'upstream' section, and I know I can't use two servers in a proxy_pass directive, but is there any way to use nginx as 'failover balancer' for two different URLs?

I'm trying to configure linux interface into trunk-only mode, that means 'accept only vlan-tagged traffic', and most importantly, ignore all untagged traffic.

I'm pretty successful at having tagged traffic to work (ip link add type vlan works nincely), but I'm having trouble to ignore all untagged traffic.

I've tried to create vlan interface with loose_binding on mode and to set main interface DOWN. It allowed me to set vlan interfaces UP, but it still show them with M-DOWN state, and none of the traffic is leaving interface (with tag or without). Basically, if I shutdown linked (main) interface, all vlans stops to work.

How to forbid linux to see traffic on 'main' (untagged, native) interface? I want to disable everything: reacting on suspicious mac addresses, ipv6, ipv4, arp, etc. I just want linux to ignore untagged traffic completely and never try to send anything untagged.

Is this possible without involving bridges and/or OVS?

Is nf_conntrack respects network namespaces? Does it creates a separate table for each network namespace based on its sysctl limits, or does it share memory (and limits) between different network namespaces?

I've searched a lot, but wasn't able to find anything definitive.

I need to save mirrored traffic for audit purposes. Traffic for audited server is send to other server. I need to capture that traffic on dedicated interface, save it to pcap files of reasonable scope (rotation by date/size), and (may be) upload and purge them.

I can glue together some bash and tcpdump inside systemd unit, but, may be, there is a ready-made solution for that?

Is there such a thing? I don't care about performance, I just need to be able to have access to the same block device from two computers. The block device does not have anything fancy (like reservations), so it's pure read/write device without caching or other nasty tricks.

I'm unable to disable JBOD mode on H330 (dell):

I've tried:

megacli   -AdpSetProp  -EnableJBOD -0 -a0

Adapter 0: Failed to Set Adapter Properties.


megacli -CfgClr  -a0


Adapter 0: Failed to Clear Configuration.

FW error description: 
  The requested command is invalid.  

I can switch H330 to RAID mode in DRAC interface (it's a dell server), but is there a way to perform this switch from within host my means of MegaCLI?

P.S.

I can enable JBOD the same way, but not to disable it:

megacli   -AdpSetProp  -EnableJBOD -1 -a0

Adapter 0: Set JBOD to Enable success.

I have following configuration: server has two network interfaces, both are attached to switch (and will become LOWER_UP after been made up). One of those interfaces will have DHCP server with a lease. Other interface will not provide me with leases.

I need such configuration for CentOS 7, that it will receive IP regardless of interface used for DHCP.

I've tried to make both interfaces BOOTPROTO="dhcp", ONBOOT="yes" and IPV4_FAILURE_FATAL="no". But there is a catch: If CentOS detects link on eno1, it will continue to try to bring up eth0, without trying DHCP on eth2. I couldn't disable eno1, because by random chance DHCP may be on eno1 (or eno2).

Is here a way to say CentOS "try to bring up all those interfaces"?

I need to see commands which send to my scsi device. I enabled scsi debug in kernel and got CDB. It looks like this: 2a 00 00 a0 09 90 00 00 08 00.

I'm searching for command line utility to decode them in human-readable format. So far I found one java-script library to do this. Is there such utility for command line?

Thanks.

I want to test mail relay. It requires me to send CRLF at the end of DATA.

I can use toggle CRLFin telnet, or -C option for NC. But is there a way to input CRLF from keyboard in linux terminal?