|
How much you looking to spend? Also AMD or intel?
|
# ? Nov 28, 2013 23:36 |
|
|
# ? May 7, 2024 19:48 |
|
As little as possible, and yes. I'm obsessed with running as many VMs as possible to make my environment and therefore the tasks, as complex as possible. Figure I will buy myself a new lab box during the holidays.
|
# ? Nov 29, 2013 02:16 |
|
MC Fruit Stripe posted:As little as possible, and yes. I'd enable small TPS first, if you aren't proc constrained, doing large JVM environments or such TPS can really help Go to host> config tab> adv settings> Mem > Mem.AllocGuestLargePage set to 0 reboot and done. (just realized I had Mem.alloGuestRemoteLargePage not sure if it means anything) Running all this on 20GB of ram most of it is TPS With large pages enabled (4MB) TPS doesn't kick in till 94~ percent, forcing small page tables (4k) which operates at the same level of tps(4k) you get better mem reclamation for a lab at the cost of some CPU overhead. Rebuilding Spring semesters coursed because gently caress all I am bored as poo poo. HA! just realized part of that CPU overhead is a vm cloning and my FREENAS VAAI trying to compensate with ~3000Mhz used Dilbert As FUCK fucked around with this message at 07:04 on Nov 29, 2013 |
# ? Nov 29, 2013 05:15 |
|
Dilbert As gently caress posted:FREENAS VAAI You can get VAAI without a "real" SAN? Tell me more. Who else does this?
|
# ? Nov 29, 2013 05:42 |
|
That's nice dilbert and I appreciate you sharing your lab but I really have no idea what you're responding to. I was asking if anyone knew of any good Black Friday deals, and I asked here because it is within the context of upgrading my home lab.
|
# ? Nov 29, 2013 05:57 |
|
MC Fruit Stripe posted:That's nice dilbert and I appreciate you sharing your lab but I really have no idea what you're responding to. I was asking if anyone knew of any good Black Friday deals, and I asked here because it is within the context of upgrading my home lab. ohh I was just trying to convey "enable this see if you still need to upgrade". newegg mas some nice DIY kits on sale evol262 posted:You can get VAAI without a "real" SAN? Tell me more. Who else does this? IIRC HP VSA and Nexenta do it. Freenas 9.1 is ISCSI VAAI only Dilbert As FUCK fucked around with this message at 06:03 on Nov 29, 2013 |
# ? Nov 29, 2013 06:01 |
|
I did go ahead and enable (or rather, disable it), because I'll take any gain I can yet. But yeah I'm definitely overreaching with my home lab. I used the word complex earlier, but it's not really complicated, it's just a lot of stuff. I've got a SQL box, well I want to stand up 2 more in a cluster and set up replication between the standalone and the cluster. I have Exchange 2010, well I want to stand up 2013, snapshot, migrate, rollback, and migrate again. I have a bunch of Solarwinds and Veeam demos I want to run, which obviously require other boxes to be up otherwise they're monitoring thin air. I want to play with the full System Center suite. I want to run a Puppet VM, GNS3, this, that, and the other. Now they obviously don't all need to be on at the same time, not even close, but you can see how it starts adding up. I don't think my lab is better, or that it gives me any particular insight, but I think you're (dilbert) coming from a strictly or overwhelming VMware point of view, whereas I've got my vSphere environment but then on top of it I'm trying to learn a little of this, a little of that. That's why I'm thinking about a second box with 64gb of memory in it, I figure between that and the existing 32gb, even I can surely run anything I'd ever want.
|
# ? Nov 29, 2013 08:13 |
|
Does someone have a good link where I can read how virtual SANs work in VMware? I need a better storage solution for my home lab (and it's black friday). The N40L I have I'm outgrowing. I was first considering a more powerful FreeNAS box with RAID 10 ZFS but then I started thinking virtual SAN might be cheaper, I just don't really know how it works. How does it protect against a host/hard drive failure, etc.
|
# ? Nov 29, 2013 14:27 |
|
MC Fruit Stripe posted:I want to run a Puppet VM, GNS3, this, that, and the other. Speaking of GNS3: buy their T-Shirt (or give them money otherwise). I already gave money, but I am also buying a t-shirt. Apparently the Canadian government is matching every donated dollar, so your money will go a long way... https://gns3.crowdhoster.com/become-an-early-release-member
|
# ? Nov 29, 2013 15:52 |
|
Dilbert As gently caress posted:I'd enable small TPS first, if you aren't proc constrained, doing large JVM environments or such TPS can really help This is actually a really cool tip for home labs that deserves more press. I wasn't aware of this functionality. From reading on it more, it looks like you lose "10-20% performance" which I assume they mean is CPU performance (sources: 1, 2)? Given a lot of environments are memory constrained instead of CPU even in production-level environments, why isn't this setting recommended more? three fucked around with this message at 00:03 on Dec 1, 2013 |
# ? Nov 30, 2013 23:59 |
|
three posted:
Yeah, you'll make ESXi a bit busier as TPS is going to attempt matching like pages. I'd like to say it really depends how much CPU you lose on a number of things like speed/CAS latency of ram, and how dynamic your environment is. Mostly if you have a somewhat static environment as per ram goes, e.g domain controllers, IIS servers, etc. You probably won't notice that much performance loss, TPS doesn't have to try and compensate for as many active memory changes. quote:Given a lot of environments are memory constrained instead of CPU even in production-level environments, why isn't this setting recommended more? I think a lot has to do that Memory is often predictable on nature as well as having a bunch of ways to reclaim memory(one of which is break up pages), you don't normally see memory spike from 5 to 20% in most environments the way a windows guest CPU may. I'm cautiously optimistic to see how it would play in a VDI environment. 10001 had a good post about it from the VM thread. 1000101 posted:Keep in mind that EPT (or AMD's RVI) doesn't necessarily mean you're using large pages; you can actually use it with small pages as well. What EPT does give you is the ability to have guest memory managed in hardware as opposed to via software. If a little extra latency doesn't hurt your apps then it may be worth forcing small pages to get higher consolidation ratios.
|
# ? Dec 1, 2013 01:08 |
|
So anyone here have experience with FreeNAS's ZFS replication and performance? Basically looking at 2 Nodes; 7x146GB 15K drives 1x400GB EFD Thinking about doing RAID-Z on the 7 146GB drives, ZIL 12GB, and with L2ARC 200GB; and supplying the FreeNAS appliance with 4vCPU's and 12GB ram with Direct IO to the RAID Controller. I'm looking to replicate data between the 2 nodes in 30-45 minute intervals. With Backups going to a VDP appliance. The concern I have is how much a performance impact on when Freenas takes the snap to when it replicates. I realize it is going to snap the changed data and the performance is relative to the amount of changed data but I was wondering if anyone has experience with it.
|
# ? Dec 3, 2013 06:29 |
|
would I be wasting my time with a couple of Cisco 871 routers to round out my CCNA lab? I've already got a 1941 and two Catalyst 2950T switches, and can borrow a 3750 from work.
|
# ? Dec 9, 2013 06:13 |
|
We just replaced some switches and now I have a HP 4000M and HP 8000M sitting under my desk. Are these of any value to me or anyone for anything or should I just junk them?
|
# ? Dec 10, 2013 06:07 |
|
EDIT* belongs in the NAS thread
BlueBlazer fucked around with this message at 07:35 on Dec 10, 2013 |
# ? Dec 10, 2013 07:16 |
|
Swink posted:We just replaced some switches and now I have a HP 4000M and HP 8000M sitting under my desk. I'd love to have an HP ProCurve for labbing stuff, since I just started a new job where we use them and I'm not that familiar with them. But holy moly those are big
|
# ? Dec 10, 2013 14:39 |
|
I'm a fan of the HP 1810 series if you want a cheap, managed gigabit switch. It's not gonna help you on anything vendor specific like the CCNA but they do support things like VLAN's, jumbo frames, link aggregation, and spanning-tree if you just want to learn the concepts (or actually want to use those features in your lab). You can get an 8 porter for under $100. Docjowles fucked around with this message at 17:37 on Dec 10, 2013 |
# ? Dec 10, 2013 17:34 |
|
Docjowles posted:I'm a fan of the HP 1810 series if you want a cheap, managed gigabit switch. It's not gonna help you on anything vendor specific like the CCNA but they do support things like VLAN's, jumbo frames, link aggregation, and spanning-tree if you just want to learn the concepts (or actually want to use those features in your lab). I'm hella familiar with Cisco stuff, I just wanted a beater HP device to test on. That looks pretty good, I'm assuming it runs the same standard HP CLI that the ProCurve devices use?
|
# ? Dec 10, 2013 18:26 |
|
QPZIL posted:I'm hella familiar with Cisco stuff, I just wanted a beater HP device to test on. That looks pretty good, I'm assuming it runs the same standard HP CLI that the ProCurve devices use? The 1810 is a webmanaged only switch I think it isn't until the 25xx or 26xx series that you get CLI management.
|
# ? Dec 10, 2013 18:27 |
|
You can grab a 2824 on ebay for 150 -- just from a cursory glance. I have it on good authority that you can replace the fans with quieter ones if you're looking to place it in your bedroom or something.
|
# ? Dec 10, 2013 18:32 |
|
If you don't care about gigabit and just want some Procurve CLI experience, the 10/100 (+ Gig uplink) 2510s (J9019B) are pretty cheap on eBay, etc. The Gig 2510s (J9279A) are still a few hundred used, but they're still an excellent value if you want a nice L2 managed switch. Edit: Wow, I might have to pick up a 2824... SamDabbers fucked around with this message at 18:39 on Dec 10, 2013 |
# ? Dec 10, 2013 18:34 |
|
Dilbert As gently caress posted:So anyone here have experience with FreeNAS's ZFS replication and performance? If you're not doing dedup you should be fine. I had a setup where I was taking snapshots on 3-4TB of real data (w/ dedup like 50TB) and there was no real performance hit even with that. Your l2arc might be a bit too big for the amount of ram you have though. I forget exactly how it goes but for each xGB of l2arc you need 1GB of ram (I think the "optimal" is 10-1) otherwise you might run into some memory performance problems.
|
# ? Dec 12, 2013 16:05 |
|
God, there's 4 threads I could post this in. This is probably the least active thread of those, but also feels like the thread where people may have run into the issue. I'm working through half a kernel of an unformed thought... I currently run a lab domain and network in VMware Workstation on a pretty beefy desktop computer. That computer is also on my regular home network. My home network is 192.168.1.x and my lab network is 192.168.10.x. The two are bridged via a pfSense VM with 2 NICs, one attached to each network. This allows the lab network to have its own environment yet also get out the internet when it needs to. I like this set up. I'm going to be standing up a second lab box. I'd like that lab box to be on the same subnet as the lab network. How am I going to do this, or what's my closest approximation? Here's a drawing which illustrates the problem I'm anticipating when I have the second lab box set up. VM1 pings VM2, but the ping reaches its first hop at pfSense, it sees a 192.168.10.x address and is like, uh that's not an IP I have information for, goodbye. Of course then we get into the option of putting lab box 1 on 192.168.10.x and lab box 2 on 192.168.20.x, but even then it feels like there's going to be problems. For example, if I want to move a VM over to the other lab box, same situation, it won't know how to route to a 192.168.20.x IP on a 192.168.10.x subnet. And even during normal course of duty, the 192.168.10.x pfSense isn't going to know where to send information for 192.168.20.x hosts. There are ways to do this, but none of them feel particularly graceful, so I'm curious how you guys would handle it.
|
# ? Dec 27, 2013 20:35 |
|
Two different physical locations is like a prime example of seperate subnets. If you move to the other area you have to change ip. Just like irl.
|
# ? Dec 27, 2013 20:54 |
|
MC Fruit Stripe posted:
Set up a static route on each pfsense vm for the other, then they will know about the networks behind each other. Make sure you turn off the firewall rule for RFC whatevernumberitwas private addresses. edit: you'll have to change ip addresses if you move vms between subnets. Alternatively get a nic for the virtual machines on each box and plug them into a cheapo router like a Mikrotik and save yourself a lot of hassle. thebigcow fucked around with this message at 20:58 on Dec 27, 2013 |
# ? Dec 27, 2013 20:55 |
|
thebigcow posted:edit: you'll have to change ip addresses if you move vms between subnets. Alternatively get a nic for the virtual machines on each box and plug them into a cheapo router like a Mikrotik and save yourself a lot of hassle. This, for what it's worth, is where I'm leaning. I'm not ready to put a rack of Cisco equipment between the two boxes to simulate separate locations, but that's going to be the end goal and another NIC for each box would need to be part of that, so I think maybe this is simply going to be a step in that direction. That plus static route might just do everything I ask of it, good show!
|
# ? Dec 27, 2013 21:11 |
|
There are two port intel nics on ebay for a reasonable price, if you can deal with the hassle this will leave room for future stupidity. If you are only going to use two computers right now you can just connect the two nics directly to each other and run a pfsense vm on one machine to handle it all. You will want to read up on vmware vswitches to understand the limitations of this but it works like hooking two switches together for almost everything. The way I have it set up under workstation 9 on windows 7 with the second nic on my motherboard is to leave it enabled in windows, disable ip 4 and 6 and the two microsoft networking things. Then use the custom network editor in workstation to bridge one of the unused vmnets to that nic. Then in the hardware settings for each vm change the network adapter to custom network segment and pick the one you just set up.
|
# ? Dec 27, 2013 21:29 |
|
I'm actually trying to do the exact same thing myself. I have the NICs but I didn't realize I could use a VM in place of a physical router.
|
# ? Dec 27, 2013 22:13 |
|
MC Fruit Stripe posted:This, for what it's worth, is where I'm leaning. I'm not ready to put a rack of Cisco equipment between the two boxes to simulate separate locations, but that's going to be the end goal and another NIC for each box would need to be part of that, so I think maybe this is simply going to be a step in that direction. Install pfsense on the 2nd lab machine. Connect the two with ipsec or openvpn. Use .10 on both.
|
# ? Dec 28, 2013 01:11 |
|
If you guys have a Micro Center close by and are looking to build a home lab for virtualization, they (microcenter) dropped the price of the FX 6300 to $90. Buy that with a motherboard for another $40 off. I walked away with three boxes (16gb ram, 8gb, 8gb) for my lab. Find some Intel gigabit NICs on ebay for $15 if you don't already have any, pick up a cheapo Dell 2708/16/24 for gigabit with VLAN support and whalla, awesome home lab. I have two Toshiba 7200RPM drives in the first box that I present to an openfiler VM for iSCSI. Benchmarking the drives in Ubuntu (using the basic disks benchmark) nets me around 100MB/sec read and 70-80MB/sec write (10MB sample). VMs are quite speedy felling, etc. Really gives a nice platform to get your VCP from.
|
# ? Dec 28, 2013 04:13 |
|
mAlfunkti0n posted:whalla
|
# ? Dec 28, 2013 22:46 |
|
Welp my CCNA is due to expire in December plus I have had a couple people tell me I need to up my credentials before I go in to interviews asking for the salary I am so its time to put together my CCNP lab. Luckily we never throw anything away and we have some recent decommissioned stuff I can use. What is not pictured is an 1841 I am going to throw in there once I can find some rack ears. I also need to get ahold of a good layer three switch so I need to scour ebay some. Syano fucked around with this message at 20:48 on Jan 7, 2014 |
# ? Jan 7, 2014 20:44 |
|
No idea where to post this, but I figure I'll take a stab here (Juniper question). So I have inherited networking at my job. I am going through and documenting the mess that our previous network guy left me. Going through our core switch clusters, I notice one site doesn't have a loopback address. From my reading, I thought Junos required this? Right now that site really isn't in production on that cluster (only one physical server running there), but it seems to be working? Someone care to set me straight on this?
|
# ? Jan 8, 2014 18:00 |
|
Moey posted:No idea where to post this, but I figure I'll take a stab here (Juniper question). JUNOS doesn't require an address on lo0, but having one certainly makes management life easier when it has one.
|
# ? Jan 8, 2014 18:02 |
|
H.R. Paperstacks posted:JUNOS doesn't require an address on lo0, but having one certainly makes management life easier when it has one. Can you expand a little more on this?
|
# ? Jan 8, 2014 18:12 |
|
Moey posted:Can you expand a little more on this? Assuming you're running an IGP like OSPF or RIP (lol), you'll always be able to reach the loopback address as long as one interface on the router is up, rather than having to try to get in using each interface address until you find one that works.
|
# ? Jan 8, 2014 19:59 |
|
Moey posted:Can you expand a little more on this? For your sites in which the clusters have a lo0.0 address, what is it being used for? Do you SSH to it? Source NTP/SYSLOG/ETC from it? Juniper FXP/ME/VME/BME ports (the physical port labeled "MGMT" on the chassis or Routing Engine) cannot be used for transient traffic, only traffic destined to/from the control plane. You can assign an IP to that interface, but you control access to the control plane via an input filter on lo0.0 even if it doesn't have an address on it. Having an address assigned to lo0.0 gives you further flexibility when it comes things like routing protocols and management of the device since your loopback addresses are available on every interface that is up/up, regardless of their assigned IP.
|
# ? Jan 8, 2014 20:07 |
|
Thanks guys. Right now these cores are not fully implemented so our inter-site routing is being handled by the old cores (OSPF). Poking around some more, he seemed to just put a random /32 address on each Lo0. Defiantly not routable anywhere on the network. Are there any recommended books either of you would advise? Each of our "core clusters" consists of a virtual chassis of 2 x SRX240 and another virtual chassis of 2 x EX4550 and 2 x EX4200. Edit: Magoo. Just started looking at what he was doing in the firewalls. I am going to have a lot of work ahead of me. Moey fucked around with this message at 21:56 on Jan 8, 2014 |
# ? Jan 8, 2014 21:49 |
|
Moey posted:Thanks guys. Right now these cores are not fully implemented so our inter-site routing is being handled by the old cores (OSPF). Book wise, you'll want to get JUNOS Enterprise Routing and JUNOS Enterprise Switching, both via O'Reilly. Those will cover 90% of what you'll be doing with the EX line. Depending on what role you are going to put the SRX's into, flow based vs packet based, there will be some overlap. There is also a JUNOS SRX Series book by O'Reilly as well, but I have not read it, the previous two I have when I was prep'ing for JNCIE. JUNOS Cookbook is decent, but also look at all the free Juniper: Day One books via https://www.juniper.net they are packed with good info and tips.
|
# ? Jan 8, 2014 22:09 |
|
|
# ? May 7, 2024 19:48 |
|
H.R. Paperstacks posted:Book wise, you'll want to get JUNOS Enterprise Routing and JUNOS Enterprise Switching, both via O'Reilly. Those will cover 90% of what you'll be doing with the EX line. Depending on what role you are going to put the SRX's into, flow based vs packet based, there will be some overlap. There is also a JUNOS SRX Series book by O'Reilly as well, but I have not read it, the previous two I have when I was prep'ing for JNCIE. Thanks, I'll grab those two books. Should keep me busy for a little while. I have been working through the Day One books already.
|
# ? Jan 8, 2014 23:14 |