zambahow can i check through esxi that the NIC has attached a SFP?
neotic_you can't, you can only see if it has link or not
danliiI have some CentOS 7 machines running on an ESXi 6.5 cluster, without "host clock sync" option, but instead running ntp clients against our in-house ntp server. The clock on the machines jumps forward two hours, give or take 2-3 minutes, so that the ntp client dies. They do this quite frequently. I don't know how to start troubleshooting this. Any ideas? :(
jzu_danlii: use host clock sync :)
danliijzu_: That's apparently not recommended by... anyone. And I fail to see why it's only CentOS machines that does this, my Debian machines are all fine.
neoticsomething wrong with the ntp client :)
celyrhmh
celyrwhat if you hwclock -r in your debian and in your centos ?
danliicelyr: hwclock -r on debian and centos gives the same result, but it differs from output from "date" on centos.
twkmsmells like a ntp implementation issue.
danliiI don't think it has to do with ntp, the drifting doesn't happen because of the ntp server reporting the wrong time, then we would have wrong time on all machines.
danliiThere seems to be something happening that moves the clock forward, and then the ntp client exits because that's how it's supposed to work, exit when it sees a drift of more than 10 minutes or so.
twkmif you say so.
hateballI recall having to use kernel params for clock drift on some suse servers, but this was on rather ancient ESX
hateballand ancient suse/kernel as well
hateballdanlii: you dont get anything like this? https://access.redhat.com/solutions/434883
danliihateball: Nope.
Churchi had one sparc server that had clock jumps. no ntp could help due irregularity. it was narrowed down to hw clock (implemented on out of band controller board) issues. i suggest to contact hw support if you still have warranty (if you run vmware on it, maybe hardware is not too ancient?)
Churchntpdate could be used to sync even if there are larger time differences, but still, most probably it's hw issue, and it's there that should be fixed, not tried to workaround with sw hacks
danliiWell, it's a virtual server, so it doesn't really have a hw clock, it just pretends it does.
danliiBut the clock is correct on the host machine and on most other guests.
Churchhmm, host itself don't have time issues?
danliiNope, none at all.
Churchis vm configured to use host's clock or has independent?
danliiIt's just certain Centos machines that suddenly jumps forward two hours in time.
danliiIndependent.
Churchtwo hours? maybe some timezone misconfig?
danliiAll our vms are set to independent, which is best practice according to vmware.
Churchand which centos versions? 4-5-6 or 7?
danliiAll our machines are configured the same way, it's just the kernel version that differs.
danliiCentOS 7.
danlii7.4 even, so it's quite up to date.
Churchcentos 7 had lot of new crap, including for time sync. all that systemd shit, time sync by default via chrony, not ntp and so on
Churchs/ntp/ntpd/
danliiHmm
danliichronyd is actually running on this particular machine as well. I hadn't noticed.
Churchone can use normal ntpd for time sync on C7. it's just not by default install
danliino, but it is installed by our ansible playbook that we run on all new machines.
danliiSo we're actually running both ntpd and chronyd at the same time.
Churchwell, google can help find how to setup/debug both time sync varieties
danliiAgainst different servers.
twkmthat seems a mistake.
twkm2 hours sounds like localtime is being used to set the clock.
danliiBut not on the next server that I check on. Of course it's not that simple.
danliiAnd the next one is off by two hours and 7 minutes.
Ussatyou should not run chronyd and ntp at same time
danliiYes, we've established that.
twkmrather chronyd and ntpd.
Ussatand TBH if its a VM you should be useing chronyd and not ntpd
Churchi'd prefer other way around. ntpd is everywhere, crossplatform/os, one majority have most experience with
Churchso i'd rather add another ansible rule to disable chronyd
Church(if on 7)
Ussatwhat you prefer and what works best are different...but its your system
Ussatfeel free to bork it up
Churchto hell innovation for innovation sake, if it isn't better in any real way and replaces for some ambitions sake what is proven to work
Ussatit is better, its quite a bit better in a VM, but go ahead
twkmcan you elucidate how it is better?
Churchhmm, can you name some things in which it is better except now having to use different command syntax for some of functionalities in it's use?
danlii*its
Ussathttps://access.redhat.com/solutions/2070363
Ussathttps://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/7/html/system_administrators_guide/ch-configuring_ntp_using_the_chrony_suite#sect-Choosing_between_NTP_daemon
UssatThats just two references, TBH use what you want, chrony is prefered
ChurchBoth links just like you just claims that it should be used (except where it cannot due limitations, lol).
Churchbut WHY?!
UssatI am not doin your research for you...I have about 1k RHEL servers in vmware and am pretty sure I know what I am doin. Do what you want
UssatThats just 2 links
twkmdissmissive, oh well.
UssatThere are lots of articles out there as to why Chrony is prefered in certian situations
Ussatit depends on his/hers setup
Church2nd link is a bit more verbose in describing differences .. but from those it seems that corner cases where chrony might be better or only choice are too rare & rather as exceptions, with ntpd serving just find most setups here.
Churchi'd probably lean to chronyd usage only in case of minimum installs where one should use as much as possible from what is there by default.
Ussatconsidering chronyd will be the default going foward
twkmso far it seems chrony was mainly created to make laptop users happier, where ntpd expects always on. i'm not sure that is compelling for most kinds of vm's -- vdi perhaps.
wahibdonI have three monitors on my host computer. When I connect to a VM with VMware Horizon and choose "full screen all monitors" it only displays on two. Anyone know why the third monitor would not be included?
danliiOk, new problem - I need to unmount a datastore from my vSphere 6.5 environment, and I can't because of some core dump files that apparently reside there. I have tried removing them with esxcli system coredump file remove -f fullpathofcoredumpfile, but it doesn't work, even with the "force" flag, it just says that the file or resource is busy. None of them are set to active. How do I proceed?
twkmturn off the vm's using them?
danliiWhich vms are that?
twkmusually one can "see" it in the client, which will show that a vm is using a datastore.
danliiUhu. But the datastore has no vms on it, and it is in maintenance mode. The dump files are from an esxi host.
danliiHm, but I could probably benefit from trying from the other hosts as well...
twkmmaybe see <https://kb.vmware.com/kb/2110152>
danliiyup, that was it. it was locked from another esxi host. silly me...
danliiNow it seems like I have a problem with locked log files.
danliiI reconfigured my hosts to log to another datastore, and reloaded the syslog daemon, but the files are still in use by a process called sfcb-svs_pgyeec - even though the file doesn't exist anymore. So what is sfcb-svs_pgyeec and how do I get it to understand that the log file it locks is not there anymore?
twkmrestart the sfcb?
danliiWhat is an sfcb?
twkmsmall footprint cim broker.
twkmlike snmp, but newer.
danliiYup, that did the trick. Thanks! :)
jaelaehmf
jaelaeole heap size issue on my 6.0 vcenter
jaelaeive run into this in the past and maybe i never fully understand what goes on here. Windows VM with vCenter. 16gb of memory. I get alerts going to yellow stating vsphere-client and vmware-dataservice-sca changed from green to yellow. KB states to increase the heap size manually. To determine the ram size run cloudvm-ram-size. But how do you properly determine what the correct manual size is.
jaelaecloudvm-ram-size -S shows tota ram allocated to each service. and the total ram on the system..
crobertscan you use an esxi host as a vagrant controller to run playbooks against other hosts
theacolytemaybe but that's a really bad idea
theacolytejust spin up a small vm
crobertsgood point'
Harm133Hey guys, any of you know what could be the reason why I can login physically on an esxi machine but not from the webinterface?
ProAliaswhat happens when you try?
Harm133just get a password denied
Harm133was on the phone with a colleague. He moved the machine from 1 subnet to another, setup a static IP in the new subnet and couldnt login anymore.
Harm133Then he changed the password from the physical machine
Harm133still no luck
Harm133then he rebooted, which solved the issue
Harm133So, thx anyways :)
mockie_Is there another way to download patches besides the my.vmware.com Java Applet portal?
twkmuse vum?
mockieI'm not familiar with VUM. It's a separate license, yes? I'm on VC 6.0, btw.
twkmit's part of the basic vcenter, but a separate install.
mockieFor both Windows client and as a VM?
mockieSorry, the documentation has my head spinning.
mockieI may have found the right place in the documentation.
FrostCyborgit's not a separate part of vCenter anymore if you're using the vCSA on 6.5 btw
FrostCyborgIt's baked in
mockieThanks FrostCyborg, I'm glad to hear that. I'm at 6.0 and not on Windows so I'm trying to figure out what to do.
FrostCyborgYeah, that alone was worth the upgrade to 6.5 for me.
FrostCyborgBut ymmv depending on environment etc.
twkma 6.5 vcsa can manage 6.0 hosts, no need to upgrade the hosts too.
mockieI figured out the download issue. There was a cookie on my computer that was jacking up the download from my.vmware.com. Only after opening a private tab could I download what I needed.
mockieI thought I read that 6.5 wasn't able to manage ESXi 5.1 (I inherited a pretty dated infrastructure).
yadnusHello, im having an issue with a p2v conversion of a win 7 machine. I created the vmdk with qemu img convert with no problems. But VMWare will not fully boot windows unless i have the HDD set to IDE ( the old machine was not IDE ) but its unusably slow. Eveyother setting i tried dumps me to a start windows repair screen. Ive done this same conversion process before with 0 issues... any ideas?