Installation problem:Vicidial-Redux-DB_x64 on LSI MegaRAID

[ronator]

Hello folks,

I tried to install vicidial-redux-DB_x64 from iso (3.0.6) on RAID 1 (LSI MegaRAID 320) but everytime I restarted, the system couldn't boot. After a while, I realized that it might be related to the fact that in yast-partitioning, I cannot choose a device-ID setup.

I tried it with a normal openSUSE (11.2-x64) and I could choose device-id (which is default) and the system booted normally.

Can someone gimme a hint, what's going wrong here ? I also saw the minor-number going up to 3.0.8 since the last week, but there's nothing in changelog related to my problem :/

thx a lot and best wishes
r0n

[williamconley]

1) we're at 3.0.8 now

2) have you also tried the 32-bit to see if the problem is there as well?

3) have you tried modifying the install after the fact to choose a device-id?

[iulianm]

I had a related problem with DELL R410 and H200 raid controller. The 32BIT version of 3.0.8 doesn't boot after installation. Tried automatic partition with volume / without ... manual ... after installation was an eror that cannot find part 2 ...

The 64bit worked for me.

[ronator]

hello there,

thanks for answering to my thread!

1) Yes, I am aware of 3.0.8 but I followed the "bug-tracking" in the related thread, and it seemed to me that there was no change related to my problem. It seems to me, it might have to do something with creating the image with SuSEstudio (because the full suse distribution boots on my RAID-system).
2) Yes, I also tried 32bit, and when entering yast-partitioner, there are two options not available: device-id and device-path ("greyed out").
3) Yes, I even tried to pre-partition the raid-system, then booting it from live-cd trying to install and walk around the partitioning-setup. But since I have no big experience what happens between live-install and first-boot, I cannot decide if those fstab-entries make sense, that are written after "os-install".
4) I'll give it a try with 3.0.8, but I don't think it could help me. I'll come back here later and post my results.

The error described by the poster before me is just the same I have: since vicidial suggests to use /dev/sda1 as swap and /dev/sda2 as / (with /boot), it hangs when rebooting, because it cannot find /dev/sda2. The fallback-option offered is senseless, because it only offers to fallback to /dev/sda2 (which is what cannot be found by the system).

5) If I cannot get the desired results, I am about to do it with a clean openSuSE-installation, but since this setup is partially like a from-scratch-concept (which I'd really love to walk around), it would not be the "golden way". Certainly, I am not sure, which perl-scripts must reside on the DB-/Web-Server and if possible I'd appreciate the live-cd-installation which could have saved me some time ... is there something else that I have to keep in mind ?


best wishes,
r0n

[ronator]

Bad news for me : the newest release 3.0.8 (64bit) shows the same behaviour as the previous version. I expected that, but since I have _no clue_ (not yet researched) what the software lacks due to the impossibility of setting a device bootable by device-id, I am still stepping in the dark. After booting the live-cd, when entering yast, I can look into the "Partitioning Summary" and the "Default Settings", where such things as ext4 are preselected. Here is also device-id selected, but it is not available when I enter section "hard drives". "lsmod" also shows the related drivers.
It's somehow confusing that the live-cd seems to recognize all my RAIDs (RAID-1 on system-disk, RAID-10 on /var and on /var/lib/mysql) but after a reboot it cannot find /dev/sda1 ... which must be connected to the (let me call it) device-id-problem, since a standard openSuSE-Installation went threw it all without any pain ...

If you wanna see any logs, I'll provide them if I can ... the system drops me to a very bare system-console (where I can call insmod but not lsmod; just as a side notice ...)

best wishes,
r0n

[williamconley]

use the live installer ... copy the fstab file. then install with vicibox, and COPY the fstab file to it.

[ronator]

personal taste may differ, but to give a little hint:

vicibox-1.0.0-RC2 did the job without any problems!

Does it sound cruel to mention the idea of setting up a vicibox-1.0.0 as DB-/Webserver and the telephony servers based on vicidial-redux ???

[ronator]

use the live installer ... copy the fstab file. then install with vicibox, and COPY the fstab file to it.

I will do anything you suggest. I used the live-installer many times and curious as I am I checked fstab after os-install ... it never showed the entires I tried to setup. I'll go and try to copy, but I am not sure if I got the idea of it. I hope I understand correctly copying the fstab from "anywhere" in the "live-system" to /etc on hdd ?!?

* press thumbs, folks ! *

[williamconley]

if the system is live you don't need to copy that file (that would be counterproductive).

if the system refuses to boot, boot from a cd and mount the drive with the "bad fstab" and copy the fstab file to it to see if you can get it to boot properly.

vicibox 1.0.0rc2 will certainly work as a db/web server, but not technically as well as a 64-bit vicibox 3.0.8 would, but if you are not expecting to fully load it (near capacity) it would be fine for quite some time. and by the time you do approach capacity, the drivers may be in place properly.

just remember to upgrade teh 1.0.0 install to match the version installed in the other servers.

[ronator]

if the system refuses to boot, boot from a cd and mount the drive with the "bad fstab" and copy the fstab file to it to see if you can get it to boot properly.

Thank you for this hint. Just to inform you: the problems seems to sit a little deeper. I installed the live-system and did not reboot. I mounted /dev/sda2 to /mnt/ and there I was looking for the content of /etc/fstab and surprisingly, everything was defined correctly (this time I re-tried it with UUID).

Code: Select all

UUID=d9b9edfd-d7b7-454e-81d7-d5690d0d8d93 swap                 swap       defaults              0 0
UUID=64716284-d936-42f4-abd3-dc32ea00f586 /                    ext3       acl,user_xattr        1 1
UUID=b0dcf1a3-2648-46b7-8674-caf23782e05e /var                 ext3       acl,user_xattr        1 2
UUID=106c07d8-e0e4-4b69-a858-0dd2fb39e57c /var/lib/mysql       ext3       acl,user_xattr        1 2
proc                 /proc                proc       defaults              0 0
sysfs                /sys                 sysfs      noauto                0 0
debugfs              /sys/kernel/debug    debugfs    noauto                0 0
devpts               /dev/pts             devpts     mode=0620,gid=5       0 0


I rebooted and the system searched itself to death, waiting for device /dev/sda2 to appear. That's the first point that irritates me: I set up the system with UUID and not device-name ?!? From where the heck does it know anything about /dev/sda2 ? Can only come from grub, doesn't it ? Well, /boot/grub/menu.lst looked "normal" ...
Now I followed your hint (got a copy of the fstab above), booted with knoppix, mounted /dev/sda1, looked for /etc/fstab ... and it was just there with all entries! So there is no need to copy. When I reboot and let me be dropped to rescue-shell (it isn't at all, is it ?) I checked the whole path /dev and there were some ttys but no block device at all. I am just speculating in the hope for a new rebuild of vici-redux: maybe PCI is initialized "too late", or maybe the drivers (megaraid_mm, megaraid_mbox or something like) are not installed correctly. You know, I don't want to bother or annoy anyone, I just really love the setup that has been created for vicidial-redux with all those install-scripts; especially when considering a multi-server / cluster-setup. I am a little afraid (or call it lazy) setting up the DB from scratch.
So if there is the slightest chance to get this problem fixed in the next update of vicidial-redux, I'd be extremely grateful.

To sum up: it may be a problem with udev or any startup-script else, maybe a "wrong" script order, but it is _not_ a bad fstab file.

Thank you very much for your time and commitment.

[williamconley]

if all you want is db, you can set up a generic 64-bit OS and increase the max db-connects to it. it does not need to have scripts running on it, or any other form of special install.

however: if you want WEB or Dialer services available on that server, you do need to install vicidial.

[Kumba]

That's weird, the FSTAB's on my SCSI raid servers don't show up like that. They look like this:

Code: Select all

/dev/disk/by-id/scsi-3600605b0001b95701445ba6ffb6e5956-part2 swap                 swap       defaults              0 0
/dev/disk/by-id/scsi-3600605b0001b95701445ba6ffb6e5956-part1 /                    ext3       acl,user_xattr        1 1


So the ViciBox Redux installer will find the drives, partition them, but upon reboot, the system says nothing is there?

I would also try running this to make sure you don't have part of some other linux distribution trying to run:
dd if=/dev/zero of=/dev/sdX bs=512 count=1
(sdX = sda, sdb, etc)

I know some distributions like Ubuntu 10, etc, use grub2 which can be a royal pain to get rid of.

[ronator]

however: if you want WEB or Dialer services available on that server, you do need to install vicidial.

I successfully installed ViciBox Redux DB x64 on a sata-hdd without RAID. First thing I noticed: the partitioner offered device-id as mount-by-option. After reboot, an automatic installation process started and after that, I had my RAID-10 (mysql) mounted and running. So I think this automatic installation process takes care of some driver-installation.

I have one question left concerning these keep-alive-scripts: vicibox-db suggested to use scripts 5,7,9 (AST_VDadapt, AST_VDauto_dial_FILL, Timeclock). Originally, I wanted to outsource the DB from our heavily loaded single-server-system. Later, I'd like to outsource the webserver on the same machine.

So, is it correct, when I want an "external"-DB-only server, I choose [X] meaning no keep-alive-script at all on that machine? And when I want to outsource the webserver, then there must run the scripts [579] ? And Do I have to stop some scripts on the phone-server (which will still act as webserver) ? Excuse those basic questions, I never had the time to learn something about multi-server setup, and now I am forced to get it running ... I'll also take a look into these vicidial-manuals as well as upgrade.txt and the from_scratch-document but as always I am very grateful if someone as qualified as you can answer my questions.

best wishes, r0n

[ronator]

So the ViciBox Redux installer will find the drives, partition them, but upon reboot, the system says nothing is there?
I know some distributions like Ubuntu 10, etc, use grub2 which can be a royal pain to get rid of.

Yes, exactly, as if after reboot the drivers could not be initialized. But the point concerning device-id still sets me up. When the hd is connect over PCI (RAID) I cannot choose to mount by device-id. If it is connected to onboard-sata-port, I can choose that option and everything works fine. See my previous posting in this thread.
Your second point also sounds interesting, but hence I partitioned and formatted the drives multiple times. Of course, I could try to "null" the hd, but as far as I know, to format a drive should make it "clean enough" to get a proper installation. Furthermore, it was never a problem with grub or MBR or something like that. Files were at the correct places and configured correctly.
From my point of view some drivers are loaded in the automatic installation process, which I couldn't reach, because the drives were not available: the typical chicken-egg-problem ...
I don't want to spam, but I think there was a difference between calling 'lsmod' in live-system compared to calling it on the installed system (with / without RAID).

Code: Select all

vicibox-db:~ # lsmod
Module                  Size  Used by
edd                    10176  0
mperf                   1523  0
loop                   17791  0
dm_mod                 85571  0
sr_mod                 16204  0
cdrom                  43051  1 sr_mod
amd64_edac_mod         28977  0
megaraid_mbox          33512  2
edac_core              49774  6 amd64_edac_mod
megaraid_mm            10393  1 megaraid_mbox
edac_mce_amd            9651  1 amd64_edac_mod
sg                     32904  0
ppdev                   9944  0
e1000                 119259  0
usb_storage            52755  0
floppy                 68253  0
k8temp                  4296  0
parport_pc             37419  0
parport                39690  2 ppdev,parport_pc
button                  6925  0
shpchp                 34692  0
pci_hotplug            31949  1 shpchp
pcspkr                  2222  0
i2c_nforce2             7561  0
ohci_hcd               37561  0
sd_mod                 41042  8
radeon                868602  1
ata_generic             3739  0
ssb                    64697  1 ohci_hcd
ttm                    64561  1 radeon
sata_nv                25493  3
mmc_core               83419  1 ssb
drm_kms_helper         32944  1 radeon
pcmcia                 66829  1 ssb
libata                211193  2 ata_generic,sata_nv
drm                   221516  3 radeon,ttm,drm_kms_helper
ehci_hcd               59876  0
scsi_mod              190568  6 sr_mod,megaraid_mbox,sg,usb_storage,sd_mod,libata
i2c_algo_bit            6728  1 radeon
pcmcia_core            23946  1 pcmcia
usbcore               229026  4 usb_storage,ohci_hcd,ehci_hcd
fan                     4559  0
thermal                20593  0
processor              45301  0
thermal_sys            17942  3 fan,thermal,processor



I could swear on the live-system there was no scsi_mod with the same dependencies, but I am not sure about that. Maybe this information can help somewhere to track down the error I faced.

Nevertheless, I do thank you for your ideas and help.

[williamconley]

I have one question left concerning these keep-alive-scripts: vicibox-db suggested to use scripts 5,7,9 (AST_VDadapt, AST_VDauto_dial_FILL, Timeclock). Originally, I wanted to outsource the DB from our heavily loaded single-server-system. Later, I'd like to outsource the webserver on the same machine.

if you read the astguiclient.conf file, it explains which keepalives should ONLY be on ONE machine. up to you which machine assuming the data in question is available to that machine (ie: the database server running a script to move recordings will likely fail since it does not have access to move recordings, but any of the servers can make changes to the database since they all have access to it ...).

Read the astguiclient.conf description and see if that answers your question.

And do it one keepalive at a time (don't say 5,7,9?, just say "5 which is the XXXX script" ... then move on to the next one, handled one at a time taking into account the data they are working with, they make sense, handled in groups it can seem too complicated)

[ronator]

Read the astguiclient.conf description and see if that answers your question.

And do it one keepalive at a time (don't say 5,7,9?, just say "5 which is the XXXX script" ... then move on to the next one, handled one at a time taking into account the data they are working with, they make sense, handled in groups it can seem too complicated)

Well, since I was a little unsure what impact install.pl could have on the system, I decided to set up the new values directly into the astguiclient.conf and restart mysqld. This file is well commented, but explanation might be a too strong phrase for this, since it says e.g. "for multi-server setup, must only run on one server". Well, that explains quite less. So, to get it started smoothly, let me see if I get the idea of it:

I should activate one script after another, check what's happening and then activate the next ? I thought it might be clearly defined which scripts must run on which machine reffering to the services that are active on it?

Test Setup:
vicibox-db (only) - pure vicidial-redux-db-x64
vicidialnow (single-server-system) - all-in-one

What I want to test is, if I can turn around the local traffic from the single-server-system to the vicibox-db. For comfortability, the webserver should remain running on the phone-server. Changing VARDBserver is not the problem, and I could see in the mysql-processlist, that cron from vicidialnow is connecting to vicibox-db. Actually, I made a test call, the vicidial-machine took it, but hung up after script agi-VDAD_all_inbound (ehm, sorry, or a similar name) was executed. I think this logically happened because I did not import the old DB in the new vicibox-DB, so there were no users, no campaigns, almost no phones. Is my assumption correct? I hoped I can test DB-connectivity without any "big moves" like mysql-dump, so that I could test if the outsourcing of mysql on a productive system is possible. And I am still not sure, which keep-alive-scripts must run on vicibox-db-only and which must run on a phone-web-server-combo ... I hoped someone could give me straight instructions ... nevertheless I'm going to reading, reading, reading ...

[williamconley]

Onward:

must only run on one server

this means: run on ONE server, any vicidial server will do, but do NOT run on any OTHER SERVER (just that one).

I should activate one script after another, check what's happening and then activate the next ?

nope. i meant ask about and discuss one at a time. otherwise the explanation and discussion will get convoluted and hard to follow

Test Setup:
vicibox-db (only) - pure vicidial-redux-db-x64
vicidialnow (single-server-system) - all-in-one

vicidialnow and viciox redux run completely different versions of vicidial ... HOWEVER: if you move the database from a "standalone" vicidial server to that pure vicidial-redux-db-x64 machine, it should work very nicely. You do not need to run ANY scripts on that machine. You DO need to TIME SYNC those two machines with each other (i do not mean have them check with each other once in a while, i mean use the ntp service to synchronize the two servers to each other permanently, one as the master)

if I can turn around the local traffic from the single-server-system to the vicibox-db

i'm not sure wht you mean by "turn around" or by "local traffic"

You may want to start here if you are clustering an old VicidialNOW installation:

http://www.poundteam.com/downloads/Vici ... v1%201.pdf

[Kumba]

Follow the steps in this thread, see if the problem persists:

http://www.vicidial.org/VICIDIALforum/v ... hp?t=14672

[ronator]

You may want to start here if you are clustering an old VicidialNOW installation:http://www.poundteam.com/downloads/Vicidial%20Multi-Server%20Manual%20by%20Poundteam%20v1%201.pdf

Awesome, thank you !
With "turn around local traffic" I just meant that data for the db is not send to localhost anymore but to an "external" host, namely vicibox. And yes, you are right, talking about one kee-alive-script at one time is a good behaviour, I'll try to "implement" it :-)

@Kumba: I happily noticed your thread and I am very grateful that this problem has been located as it might even help others who could have faced the same problem!

@Bill: Yes, after some try outs with those two machines I decided to switch off all scripts on all servers, and ... the call was send to the in-group and logged to vicidial_closer_log. Of course, deactivating all those keep-alive-scripts is bloating my mailbox with messages like this:

Code: Select all

unknown remote host:  at /usr/share/astguiclient/AST_vm_update.pl line 174

But I feel optimistic that (step-by-step) I'll get it doing what I want, considering the very valuable help this community and "its founders" offer.

I cannot express my gratitude in words,
best wishes,
r0n

[williamconley]

Code: Select all

ask about and discuss one at a time.

i don't see how that translated to "TURN THEM ALL OFF"?

[ronator]

hello again,

good news and thank your for the link to that ViciDial-Multi-Server-Manual. I should shame on myself, cause I saw this document once and then later I forgot it. But this really helped me out, it confirmed some of my assumptions (cron & mysql) and it gave me some important information I was not sure about previously. So, thanks again for that.

What you quoted in the last post: I misunderstood you. But maybe u got another great link to a document, where all those keep-alive scripts and their functions are explained altogether ? I'm working towards the habit of reading some of those nice docs that are always shipped with vicidial (like load-balancing.txt).
So, right now my test-vicidialnow (which is a copy of the productiv-system) is connected to the vicibox-DB. No more warnings from cron per email, calls are obviously logged to the DB. To me it seems, it's working!
Gotta remind myself not to forget to check Kumbas suggestions!

What I reached is just to prove that we can "outsource" the DB from the ex-single-server. But I wanna go further: 4 phone-servers + 1 DB-Web-Combo. Since I definitely will use those USB-timing-sticks to get rid of ztdummy, it can become tricky to set up, can't it ? Ah, only the phone-servers need them, or am I wrong? Furthermore I will switch to vicidial (redux) for the phone servers, but I still had no time to think about how I could migrate the vicidialnow-system. You may correct me, but a set up like 4 phone-servers (3 vicidial-redux + one vicidialnow): is this possible (with same agc-version) or is it not recommended ? And if not, how can I take care for all my "custom dialplan entries" that they still work on vicidial-redux ?

Thanx for your support so far!

[ronator]

hello everybody,

I succeeded in setting up a multiserver system (1 DB + 2 phone-servers, one of them with web on it). Because I am a curious and sometimes a crazy guy, I wanted to see if I can get vicidialnow 1.3 (VERSION: 2.2.1-237
BUILD: 100510-2015) being nice to vicibox-redux 3.0.6 (VERSION: 2.2.1-237
BUILD: 100510-2015) and with the vicibox64-DB. I followed the Load-Balancing.txt and made all the changes where they were needed. And, wow, they registered to each other and I could connect an agent from vicidialnow and another agent from vicibox. What irritates me, is that I had to define the vicidial_conferences to match the server ip. I thought, Load Balancing also works this way, that when server 1 is "full", the agent would be logged in seaminglessly to another server on the cluster. Have I understand this wrong ? Ok, the console of the vicibox-server drops an error:

Code: Select all

[Dec  9 13:05:01] ERROR[8805]: utils.c:966 ast_carefulwrite: write() returned error: Connection reset by peer
[Dec  9 13:05:01] ERROR[8805]: utils.c:966 ast_carefulwrite: write() returned error: Broken pipe

I searched for this, found an old thread in here from 2009 referring to asterisk 1.4.21, where there was a discussion that this is not a real error. Since I have asterisk 1.4.27-1 running on the vicibox, I wonder if I can ignore this error as well ? Actually, I get no systemmail or warnings per cron, so I believe it' all fine, but I do not know ...

I will now add another fresh vicibox to the "cluster", let's see, if it also drops errors ...

[ronator]

is it correct to say, that those two keep-alive scripts ( in a multi-server-setup ) ,which should be running only on one host (say vici-1)

# 5 - AST_VDadapt
# 7 - AST_VDauto_dial_FILL

that they are in charge for "relaying"/forwarding calls to a campaign across all servers ? So, I have to forward all calls to vici-1 and this machine will take care of forwarding to ready agents on the other servers ?

[mflorell]

No, those processes do not touch active calls. The FILL script will place additional calls if there is a SHORT condition. The Adapt script is the predictive algorithm as well as a stats generator.